http://eiop.or.at/eiop/texte/1999-002a.htm
:
27.1.1999| This paper's comments page | Send your comment! to this paper |
| The Determinants of Control of Commission’s Executive Functions(*) |
| Fabio Franchino |
| European Integration online Papers (EIoP) Vol.
3 (1999) N° 2; http://eiop.or.at/eiop/texte/1999-002a.htm |
| Date of Publication in :
27.1.1999 |
| | Abstract |
Back to homepage |
PDF | PS
| | This paper's comments page | Send your comment! to this paper | |
1
With the emergence of the European Union as a type of ‘regulatory state’ and of the Commission as a regulator (Majone, 1994; Majone, 1996: 61-79), the control issue has become central in the study of the Union’s activities. As Majone (1996) puts it, the demand for regulation, being it the result of a functional spillover or a redistributive bargain, is matched on the supply side by the delegation of policy-making functions to supranational institutions. The main beneficiary of such delegation is certainly the Commission and, with delegation, comes control.
In this paper I focus on questions of Commission accountability and mechanisms of Member States control, with particular emphasis on the system of committee control termed comitology. The literature on control, and specifically committee control, in the Community is slowly emerging as a topic of study in itself. For a large part, it is simply one aspect in the general description of the activities of the Commission, the Council of Ministers and the Member States (e.g. Dietrich and Wessels, 1996; Docksey and Williams, 1994; Edwards and Spence, 1994; Hayes-Renshaw and Wallace, 1997). Other students referred to it as the object of contention between institutions in the co-operation procedure (Fitzmaurice, 1988). Bradley (1992) has been one of the first legal scholar to publish on the subject, and he has been followed by a series of in depth qualitative studies with strong legal flair (Bradley, 1997; Joerges and Neyer, 1997; Vos, 1997). In political science, Majone (1994) and, especially, Pollack (1997) have framed the issue of control within the general literature on regulation, delegation and agency. Formal works have assessed the impact of committee control on the balance of power between the Commission and the Council (Steunenberg et al., 1996b) and in case of parliamentary involvement (Steunenberg et al., 1996a).
However, with few exceptions, there has been a distinctive lack of hypothesis testing and, in my opinion, even these contributions fail to clearly specify the causal link between independent variables (whatever they might be) and dependent variable, namely the establishment of committee control and the stringency of control. Joerges and Neyer’s (1997) work is probably the best example. Their in-depth qualitative study based on questionnaires with members of the standing committee for foodstuffs provides valuable insights on the working of such committee. However, on the one side there is a heavy reliance on the operative aspects of the committee activity (rather than on the strategic/informative implications of its establishment). On the other side, there is an abundance of independent variables and theoretical frameworks (e.g. functional/legal needs, institutional balance, intergovernmentalism and, even, constructivism).(1) Dogan (1997) has published one of the few quantitative studies (see also Institut für Europäische Politik, 1989) and he has fallen in a similar trap. He assesses in details, although only descriptively, the preferment toward control of the Community institutions, then he suggests a series of interpretative keys varying from neofuctionalism, neorealism to administrative convergence and new public management. Finally, Wessels (1998) sets out a specific hypothesis testing exercise with, unfortunately, several flaws. As for the dependent variable, he interchangeably uses implementation control committee and other types of committees, for which then he does not seem to provide original data to substantiate his ‘fusion’ theory from his previous contribution (Wessels, 1997). I also find difficult to spell out the implications of contending theories he suggests, some of which are normative (e.g. the federalist one) while others are simply incorrect.(2)
2
In this article I limit the analysis of control and comitology to one theoretical framework, namely agency theory, whose hypotheses I then quantitatively test on a sample of Community secondary legislation. The article is divided in five sections. In the first two I review the political science literature on delegation and control and apply the control side of agency theory to the activities of the Commission. This part relies on Kiewiet and McCubbins’ (1991: 5) work on delegation in the American Congress and Pollack’s (1997) application to the Union institutions. The analysis is however more specific and focuses solely on the Commission. Then, after a description of the committee system and brief descriptive statistics in the third section, I tests McCubbins and Page’s (1987) propositions on the determinants of control in a stratified sample of secondary legislation adopted between 1987 and 1998 in the fourth part of the article. The results show that unanimity rule, conflict between Community institutions and uncertainty are key determinants for the establishment of ex post procedural control of Commission’s implementation activities. Conflict and uncertainty are also important factors affecting the degree of stringency in control. The article concludes with some comments on the impact of these procedures on Commission’s autonomy, with reference to other works on the subject.
Whatever the reason for delegating policy-making authority to the agent, the preferences of principals can systematically differ from those of the agent with the risk of the latter pursuing its own interest at the expense of that of the principals. Kiewiet and McCubbins (1991: 5) stress that ‘there is always some conflict between the interests of those who delegate authority (principals) and the agents to whom they delegate it. Agents behave opportunistically, pursuing their own interest subject only to the constraints imposed by the relationship with the principal. The opportunism that generates agency losses is a ubiquitous feature of the human experience.’
The cost of this opportunism, alternatively termed shirking or bureaucratic drift, is coupled with a second process, known as slippage, when the agency design itself is an incentive for the agent to behave in ways that are costly for the principals (McCubbins and Page, 1987: 411). In the institutional framework of the Community, agency losses can be generated not only when Commission’s preferences differ from Member States’ or Parliament’s (shirking) but also because the Commission has the monopoly of legislative initiation that can be used to pursue its interest (slippage). Crombez (1996: 213) points out how this design gives the Commission negative power, that is the ability ‘to maintain the status quo even though a qualified majority in the Council prefers to change it.’(3)
In the face of these potential losses, principals can adopt a set of measures of containment. Kiewiet and McCubbins (1991) list four classes of such measures. The first concerns with the ex-ante design of the agency. The economic literature focuses mainly on the participation constraint that an agency contract must satisfy and the related issue of specifying compensation schedules and sanctioning mechanisms (Holmström, 1979; Kiewiet and McCubbins, 1991; Pratt and Zeckhauser, 1985). In political science the stress is on those structural arrangements that define the scope and the domain of regulatory targets, the legal instruments available to the agent to implement the act and the administrative procedures required to use the instruments.
3
The second set of measures available to limit losses includes screening and selection mechanisms. In the private sector this concerns with the investment entailed in hiring to avoid adverse selection and the signaling process to bridge the informational gap about abilities and preferences that is traditionally present between potential principals and agents. Appointment and signaling is equally important in politics. Procedures allocate appointment powers to political actors, define the timing of the appointment and the formal requirements of candidates. Informal requirements, such as those geographically, ethnically or seniority related, are also ubiquitous.
Third, the principals can monitor and influence agent’s behavior ex-post by establishing monitoring and reporting requirements. There is ample evidence of such requirements in the private sector whereby practically every employee needs to fill out progress reports for their own supervisor. In the legislation analyzed for this article about 60 percent of the sampled acts require some sort of exchange of information between the Commission and other actors. The problem with reporting is untruthful revelation if agent and principal’s incentives are incompatible. The agent is tempted to strategically reveal information so that his or her activity is seen under a favorable light. McCubbins and Schwartz (1984) point out that principals might want to offset this problem by supplementing reporting requirements with oversight mechanisms of two types. First, they might establish ‘police patrol’ oversight comprising audits, investigations, public hearings, field observations and other form of direct monitoring. These methods of control require a great deal of investment of resources in term of time and effort and, as Pollack (1997: 111) has noticed, is likely to be undersupplied in case of multiple principals because of the public nature of the control good.(4) A less costly and centralized way of control is obtaining information from affected third parties such as citizens or interest groups. McCubbins and Schwartz have dubbed this second type of oversight mechanism ‘fire-alarms’. They operate via the establishment of rules and procedures that enable third party to monitor and redress administrative decisions. They also have the advantages of focusing on violations that are important for relevant political constituencies, of solving problems of contract incompleteness and of allocating monitoring costs to third parties. However, all this with the caveat that this type of oversight is likely to be biased in favor of vigilant and resourceful groups or individuals (Moe, 1987; Pollack, 1997). This article investigates the determinants of ex-post control of the ‘policy patrols’ type on the Commission implementation activity.
The final set of measures at the disposal of principals to limit agency losses rely similarly on a third party, but this time it is explicitly established by the principal. These are termed institutional checks and ‘require that when authority has been delegated to an agent, there is at least one other agent with the authority to veto or to block the actions of that agent’ (Kiewiet and McCubbins, 1991: 34). The control takes place because agents are set to have conflicting incentives and different reporting requirements to the principal.
To conclude, it is unlikely that only one type of these measures is adopted. Hood (1991) observes how, especially in highly technical domains such as regulation, we are more likely to observe pattern of ‘interpolable balance’ that is a network of complementary and overlapping control and self-policing mechanisms to ensure that the system is under control. The thrust of this article is hence looking at only one type of control, that is committee control, recognizing that it is embedded into the richer and more complex framework of systemic Community control as will be briefly considered in the next section.
4
In this section I will sketch out the mechanisms established to control Commission’s activity using the fourfold categorization suggested by Kiewiet and McCubbins. I will rely on Pollack’s (1997) contribution on delegation and agency in the European Community to comment on their effectiveness and credibility.
The scope of the activities delegated to the Commission by the Treaty has been relatively broad. Article 155 (EC) specifies that the Commission shall ensure the proper functioning of the common market and the application of Treaty provisions. The Commission has also Treaty-based executive powers in the field of competition, safeguard measures and the administration of structural funds and can also pass legislation in few areas.(5) However, the bulk of legislative and executive powers are conferred upon the Commission by secondary legislation including administrative and regulatory functions. Further, a rather broad range of instruments is at the disposal of the Commission. These includes the power to initiate legislation and infringements proceedings, to take decisions, to formulate recommendations and to deliver opinions under the general procedural requirements set up by the Treaty.(6) The ex-ante design of the agency mandate for the Commission is more a vehicle for conferring relatively broad policymaking functions rather than controlling agent’s activities and containing agency losses. There are however two important exceptions to this rule, namely the two pillars of the Maastricht Treaty, the Common Foreign and Security Policy and the Cooperation in the Field of Justice and Home Affairs. Commission’s powers are heavily curtailed in these policy areas. There are only provisions for the Commission to be associated with the work carried out in these fields and the initiation power is shared with Member States,(7) the Commission loses also an important ally in the European Court of Justice whose judicial review has been excluded in these areas. Although there has been no revision of the agent’s original mandate, the Maastricht Treaty has been the first example where agency design has been explicitly used to limit the risk of agency losses.(8)
A second mechanism to ensure Commission’s accountability is through screening and selection mechanisms. According to article 158 (EC), Member States, after consulting the Parliament,(9) nominate by common accord the President and the Commissioners. The appointment is carried out after the approval by the Parliament of the Commission as a body. Thus, the possibility of selecting their preferred agents varies across Member States and the Parliament, the latter is for instance more successful in affecting the nomination of the President rather than that of a single Commissioner. Further, using dismissal powers to sanction agent’s behavior is scarcely credible and institutionally costly. The Court will dismiss individual members of the Commission on an application by the Council and the Commission only if he or she no longer fulfills the conditions required for holding the post and in case of serious misconduct.(10) Such case is highly improbable, as it is unlikely that the Parliament would approve a motion to censure the whole Commission as from article 144 EC. Finally, Commission members must comply with general requirements but these can be scarcely considered effective in controlling agent’s preferences.(11) Since censure and dismissal are non-credible and institutionally costly, the Commission is effectively in office for five years. Each Member State can only use, at the end of the term, their reappointment power of one or two commissioners and (shared with the Parliament) of the President.
5
The risk of agency losses is hardly eliminated through the appointment process. Instead, more successful in controlling Commission’s activities are direct oversight mechanisms. The following sections of the article are dedicated to the determinants of these measures of control. Here, to complete the general picture, I will briefly analyze the remaining measures Kiewiet and McCubbins consider at the disposal of principals to limit agency losses, that is institutional checks and fire-alarm oversight. Pollack (1997: 116) observes that ‘almost every EC institution besides the Commission plays a role in monitoring and checking the Commission’s behavior’. The European Court of Justice reviews the legality of Commission’s acts and other Community institutions can bring actions to the Court on ground of lack of competence, misuse of power and other infringements (including inaction). The Court of Auditors examines the reliability of the accounts and the legality and regularity of financial transactions carried out by the Commission. It produces annual audit reports and special reports at the request of Community institutions.(12) Finally, as noted by Pollack, articles 173 and 184, by allowing natural and legal persons affected by a Commission’s decision institute proceedings against such decision, effectively set up a system of fire-alarm oversight, thus delegating to third parties the control of the agent.
These mechanisms are complementary and overlapping but they still cover a small subset of Commission’s activities. The great bulk of Commission’s legitimate areas of intervention has a regulatory character and financial considerations play a considerably lessen role. Of the sample of legislation I have analyzed for this article, a mere three percent contains financial provisions of some relevance such as the allocation of structural funds and environmental protection projects. The system of decentralized control envisaged in articles 173 to 184, originally designed to question Commission’s actions and Community law, has become, through a series of decisions of the European Court of Justice, an enforcement mechanism to allow individuals to challenge national law if incompatible with European law. Similarly, the Court has radically restricted the circumstances under which individuals can proceed against Community actions, thus limiting the chances of declaring them ultra vires (Alter, 1998; Burley and Mattli, 1993). Member States then have to resort to a much more intrusive and costly form of oversight that directly focus on the regulatory activity of the Commission and cannot rely on the biased and restricted control of other institutions or private parties. It is this type oversight that takes the form of implementation committees that the following sections of the article consider.
The control of the implementation activity of the Commission by Member States has been essentially carried out since the establishment of the European Economic Community in 1957. Initially, though, it was on a rather ad–hoc basis and generally predominant in the Common Agricultural Policy since it was the first centrally financed and directed policy domain of the Community. The first price support policies and legislation of Community preference in agriculture established also the first oversight committees in the form of a management committee procedure. As the areas of intervention of Community legislation expanded, so did the variety of control procedures (Bradley, 1992; Demmke et al., 1996; Vos, 1997). On the first of July 1968 the customs duties amongst Member States were lifted and, in the same year, the regulatory committee procedure has been established in the field of customs law. Also in 1968 the Legal Affairs Committee of the European Parliament had designed the threefold classification of implementation committee that, albeit modified, is still present today (European Parliament, 1968-69).
6
It was however the Single Market initiative with its considerable legislative burden for European standardization,(13) that gave the impetus to the Council to reorganize the supervisory procedures of Commission’s implementing powers, twelve days only after the Single European Act entered into force on the first of July 1987.
Council Decision 373 of July 1987 rationalized this system of control and specified four main types of committee procedures: advisory, management, regulatory and safeguard. With the latter three having two variants each, the total number of distinct procedures amounts to seven. With the exception of the advisory and safeguard procedures, the control of Commission’s implementing legislation is two-tiered: the relevant committee oversees the act in question first, then it might refer it to the Council of Ministers. In the safeguard procedures, the first tier is bypassed and the act goes directly to the Council of Ministers for evaluation, in the advisory procedure there is no referral. When the Council is involved, the decision rule to adopt Commission’s measure is qualified majority.
These committees are composed of permanent representatives of the Member States. Generally, representatives of the rank of ambassadors are unlikely to seat in a committee because the large majority of the legislation under scrutiny is of technical nature; deputies with the rank of minister plenipotentiary or officials from national technical ministries take their place. (Docksey and Williams, 1994: 121-125). Committees are chaired by a senior Commission official, normally at director level, who controls the agenda and submits to the committee the implementing measures for its consideration, subject to a deadline set by the chair. The chairperson has no vote in the deliberation of the committee.
In the advisory committee procedure I, national experts issue an opinion before the Commission implements the measure. The Commission is requested to take the utmost account of such opinion but, if it chooses to disregard it, there is no referral to the Council. There are ways Member States can influence Commission’s activity by forcing a vote by simple majority in the committee, requesting to have their minority position recorded in the minutes or insisting that the Commission report how it has taken account of committee’s opinion. However, this procedure provides the Commission with the greatest autonomy and Member States’ influence over its decision-making powers is relatively limited.
In the following four procedures, national experts act as gatekeepers. In the management committee procedures IIa and IIb, the committee decides by qualified majority whether to submit the draft measure to the Council. In case of inaction or favorable opinion, the Commission may adopt the measure with immediate effect. If the committee decides to refer the measure to the Council, there are two procedural variants that differ on the timing of Council control. In variant a, the Council deliberates after the measure is applied although the Commission may decide to defer implementation for a maximum period of one month. In variant b, Council control takes place before adoption because the Commission must defer implementation for a maximum of three months. In both variants, if the Council does not act the default is the measure proposed by the Commission.
In the regulatory committee procedures IIIa and IIIb, the committee decides by qualified majority whether not to submit the implementing act to the Council. If such majority is not reached or the committee does not deliver an opinion, the measure is deferred and submitted the Council. The two variants that follow differ with regard to the default condition. In variant a, if the Council does not act the proposed measure shall be adopted by the Commission. In variant b, inaction leads to a similar outcome only if a simple majority in the Council does not object. In such case, the status quo ante prevails.
7
Finally, the safeguard committee procedures IVa and IVb do not require the establishment of a committee of national experts. The Commission must notify directly the Council prior to the adoption of a safeguard measure and any Member State may refer the Commission’s decision to the Council, which can revoke, modify or confirm the measure within a set time limit. Similarly to the regulatory procedures, the two variants differ with regard to the default condition. In variant a, if the Council does not act the proposed measure is adopted by the Commission. In variant b, inaction revokes the measure. Secondary legislation may amend these procedural requirements especially with respect to variant a, it is common that enabling legislation provides for Council control to take place after the Commission adopts the implementing measure.
Figure 1 summarizes the key features of these implementation committees.
The procedures described can be arrayed along three dimensions with respect to role that the Council plays in controlling Commission’s activities. These are 1) the decision rule in the committee to refer measures to the Council, 2) the timing of Council control and 3) the default condition if the Council does not act. This classification will be used in the following sections to develop an index of stringency of implementation control.
As mentioned in the introduction, there are not many studies on the incidence of comitology in the Community policies. To my knowledge, the report by the Institut für Europäische Politik (1989), the edited book by Pedler and Schaefer(1996) and Dogan’s (1997) article are the first academic and quantitative works which has been carried out in this field. In this section I compare Dogan’s results with those that emerge from my data set. This comparison is partial because the criteria of data selection differ;(14) nonetheless it provides interesting confirming and disconfirming evidence at least on a descriptive basis. More rigorous inferential analysis will follow.
Dogan observes that comitology procedures have been used in about 20 percent of all Council legislation enacted since 1987 and points out a consistent longitudinal trend towards more control. He founds out high incidence in company law, financial services, justice and home affairs, veterinary control, followed by customs, transport, health, food and development aid, while lower incidence in welfare, regional and competition policy, industrial adjustment, education and employment, taxation and procurement.
Figure 2 shows the incidence of comitology procedures in the different common policies in non-amending secondary legislation adopted since 1987.
8
More than 35 percent of this set of legislation has some sort of procedural control lending some credit to the thesis of increasing use of comitology. In some policy areas there are too few new legislative acts, making interpretation inadvisable.(15) By contrast, in four areas more than 50 percent of new legislation has comitology procedures, these are social policy, environment, approximation of laws and transport. Further, these areas show also a higher incidence of more restrictive procedures,(16) 100, 60, 57 and 74 percent respectively of all procedures are of the most restrictive types (i.e. regulatory and safeguard). At least for non-amending legislation, this seems to disconfirm Dogan’s (1997: 41) conclusion that high level of comitology is associated with low levels of restrictive comitology. In effect, in areas where the incidence of control is medium (agriculture and free movement) or low (commercial policy and customs unions), the percentage of restrictive procedures are also relatively low (28, 50, 33, 30 percent respectively). The sectoral patterns identified by Dogan are however confirmed, environment, approximation of laws, transport, agriculture and free movement are the areas where committee control is most extensively used.(17)
McCubbins and Page (1987) formulate two general factors that explain the establishment of control procedures, namely uncertainty and conflict. Uncertainty affects the distribution of information at the expenses of legislators who find difficult to discern the optimum policy actions and, probably, also their ultimate interests. Uncertainty also increases the need for information but also the cost to retrieve and process it, legislators would then prefer to delegate regulatory choices and instruments to the agent, with the attached information costs, and ‘sit back in an oversight role awaiting clarification of the issue’ (McCubbins and Page, 1987: 417). The procedural requirements become more restrictive for two reasons. First, the need for legislative control increases as the domain of scope and instruments delegated to the agent broadens. Second, the political risks attached to different regulatory alternatives increase with uncertainty. It is less clear which policy strategy is the most appropriate and the preservation of the status quo becomes relatively more important. Thus, legislators establish more stringent procedures to make this choice more difficult.
Increased conflict among legislators leads also to more confining procedures. McCubbins and Page’s line of reasoning is as follows. Conflict makes harder for a decisive coalition of legislators to narrow down the range of policy making functions to be delegated to the agent because the exclusion of some issues may lead to the break down of the coalition. Controversial aspects about implementation are hence deferred after the writing of the legislation and the agent’s mandate remains rather large. There is then incentive to control agent’s behavior ex post. Further, political risks of taking alternatives decisions increase with conflict and therefore there is more need to direct the agent through procedural requirements.
9
Uncertainty is more than policy specific, it is issue specific. In commercial policy, it is easier to set import duties of the common customs tariff than designing an anti-dumping regime or legislation against illicit commercial practices. In agriculture, fixing intervention prices for some products is less complex that establishing guidance funds and support systems for farmers. I would argue that an acceptable way to quantitatively operationalize information asymmetry is using the word count of the specific legislation excluding annexes, tables and recitals.(18) It seemed to me that the length of the legislative act and, more specifically, the word count reflect the complexity of the issue and assure objective cross-policy and cross-issue comparability. In relatively simple legislation such as setting duties, prices and import quotas, the word count amounts to less than one hundred words. More complex acts on import surveillance or technical directives on environmental policies may require from five hundred to over a thousand words. We should expect an increase in the length of the legal text to increase the odds of having some sort of ex post procedural control as well as to increase the stringency of control.
Conflict between legislators has been operationalized using two variables: 1) the number of amendments that the Council approves over the Commission’s proposals and 2) the number of rejected parliamentary amendments. The first variable measures the conflict between the Council and the Commission.(19) Although we cannot attach the traditional label of legislator to the Commission, having only the monopoly of proposal but not veto power, this dimension of conflict is essential in the Community. In formulating their hypotheses, McCubbins and Page disregard the role and the preferences of the agent because of the flexibility with which American legislators can establish and dismantle agencies and because the latter have no legislative or executive role. In the European Community there is no such flexibility for legislators because the agent, that is the Commission, is a hybrid body in classical constitutional terms. It carries out traditional administrative functions, frequently shared with national administrations, but has also to provide executive leadership and legislative gatekeeping (Lenaerts, 1991). On average, there are two Council amendments per act in the sampled legislation,(20) but the variance (twelve) is relatively large. This is because, in few cases, the Council has introduced more than ten amendments.
The second measure of conflict, rejected parliamentary amendments, gauges the intensity of conflict between the Parliament on the one side and the Commission and the Council on the other. The relevance of this variable has been tested on a subset of cases where the Treaty provides for either a parliamentary opinion or a vote. About 45 percent of the sampled legislation fall under this category. On average, less than two parliamentary amendments per act have been rejected in the sampled legislation, but the variance (fourteen) is even larger than with Council amendments. This is probably because parliamentary activism is related to institutional resources (Scully, 1997). Parliamentarians are more likely to table amendments under those procedures where the Parliament can effectively influence the final policy outcome.(21) We will focus in more details below on the interaction between legislative procedures and parliamentary amendments and their likely impact on control procedures. To sum up, an increase in the number of either type of amendments should increase the odds of procedural control and of stringency of control.
10
A third categorical variable, namely legislative procedures, has also been used in the analysis. This is coded as an indicator using qualified majority as the reference category and unanimity and the procedures where there is a parliamentary vote (i.e. co-operation and co-decision) as the comparing categories.(22) In his empirical analysis, Dogan (1997) sees an increasing correlation between control procedures and qualified majority. He, however, bases his comment on descriptive statistics rather than hypothesis testing. His implicit(23) line of reasoning is as follows. Since the move from (formal or informal) unanimity to qualified majority implies a loss of sovereignty for Member states, they are more likely to impose procedural control on Commission’s activity. This view runs counter to McCubbins and Page’s argument in two ways. First, the permanence of unanimity in the Treaty is a sign of conflict among Member State about the substantive content of common policies. Second, under unanimity the Commission needs to gain support for legislative intervention from the Member State whose preferences are the farthest away from the Commission’s. Contrary to Dogan, both factors lead us to expect that unanimity should be positively related to control and control stringency. While, when the Parliament is involved in a legislative procedure, we cannot predict, in principle, a clear direction of its impact on control because it depends on its preferences vis-à-vis other Community institutions and on whether the resources provided by the procedures allow it to affect the policy outcome.(24) Empirical studies stress the strong opposition of the Parliament to restrictive control procedures (Bradley, 1997; Dogan, 1997) especially because the Parliament is not involved in such committees. For the time being, however, we refrain from formulating specific hypothesis and the issue will be dealt in greater details below.
A set of other variables has also been used to operationalize uncertainty and conflict,(25) however those selected seem to have the strongest theoretical justification, allow some degree analytical separation between conflict and uncertainty and minimize, albeit not sufficiently, the problem of collinearity.
These hypotheses have been tested on a stratified sample of non-amending secondary legislation passed between the first of July 1987 (when the Single European Act came into effect and the Council decision reorganizing committee procedures was approved) and the first of October 1998.(26) The legislation is non-amending because we need to control for the position of the status quo and I would contend that an appropriate and efficient way of doing this is by selecting only the first legislative intervention in a policy issue. The population includes 1372 directives and regulations and its characteristics have already discussed above.
To test McCubbins and Page’s propositions we could estimate a log-linear model but, since some of the independent variables used are at the interval-level, we would loose precious information in an eventual re-coding.(27) I instead employ two complementary strategies. The first consists of running a series of binomial logistic regressions that compute the odds that a specific procedure is introduced in an act using as baseline the cases where there are no such control procedures. The second develops an index of stringency of implementation control from the committee procedures and employs cumulative logits to estimate a general model of ex post control of the Commission.
11
Although I consider the selected measures of conflict and uncertainty the most appropriate to test the hypotheses, problems of collinearity are unfortunately still present.(28) There seems to be a positive association between the complexity of an issue and the degree of conflict between actors. A way to deal with this problem is to estimate models that include different independent variables. Figure 3 shows the coefficients in a series of binomial regressions in two models. The first focuses mainly on conflict, which is operationalized with the categorical variable on legislative procedures and the number of Council amendments. The second retains the procedural variable and substitutes Council amendments with uncertainty (i.e. word count).
The models perform rather well since, generally, more than 90 percent of cases are correctly predicted. However, the discrepancy between the goodness-of-fit and the log-likelihood ratio invites caution in the interpretation of the models, especially for the advisory and safeguard procedures where sensitivity to departures from model fit is substantially reduced.(29) The bad performance of advisory is probably due to the fact that this is a very permissive control procedure; independent variables are less relevant because the difference with the baseline category (no control) is minimal.
Conflict, operationalized as number of Council amendments, is consistently the most significant determinant in affecting the probability of some kind of procedural control. With the exception of the advisory procedure, when the number of Council amendments increases from zero to two, the odds of procedural control increases, on average, by a factor of two (ceteris paribus). That is, the probability of ex post control increases by more than two percent (more than four in case of the regulatory procedure). If we move along the whole spectrum of values that this variable takes, it is almost certain that we will have some sort of ex post control. With an increase from zero to 16, we have an increase of a hefty 93 percent in the probability of having a regulatory committee, 84 percent for the safeguard and 66 for the management committee.(30)
Also the models incorporating uncertainty perform well although, at least in term of statistical significance, somewhat less convincingly than conflict. Except for advisory, when the length of the act increases by five hundred words (say, from an act setting a customs tariff to one administrating a quota), the odds of procedural control increases by a factor of three (ceteris paribus). The probability that there will be some sort of procedural control increases by more than three percent (almost five in case of the regulatory procedure). If there is a need to adopt complex environmental legislation (say, with an increase of two thousand words), the probability of having control to no control increases by 22 percent for the management committee, 66 for regulatory and 75 for safeguard.(31)
Finally, at least for the management procedure and, partially, for the regulatory one, the proposition that unanimity is a sign of conflict among Member States or that it structurally induces more control seems validated. Ceteris paribus, the use of unanimity compared to qualified majority increases by more than 40 percent the chance of ex post control in the form of a management committee (more than twenty percent for regulatory).(32) More difficult to interpret is the result from the second variable of legislative procedures. Although only for the regulatory committee, the presence of a parliamentary vote increases the probability of this type of control by more than 35 compared to qualified majority (ceteris paribus). This result seems to be at odd with the empirical evidence (Bradley, 1992; Bradley, 1997; Dogan, 1997) and the institutional preferences of the Parliament. This however has not to be the case. There is in fact evidence that the relation between control and parliamentary role is spurious because this institution is likely to be involved in legislation that has an above average level of conflict and uncertainty.(33)
12
In effect, McCubbins and Page’s propositions are confirmed if we look at the subset of cases where the Treaty provides for a parliamentary opinion or vote. Figure 4 shows the coefficients of the binomial regressions for the management and regulatory committees. Here, the number of rejected parliamentary amendments substitutes, as a measure of conflict, Council amendments. While the procedural variable is a dummy taking 1 for unanimity and 0 for qualified majority.(34)
Also in this test models perform well, with almost 80 and more than 90 percent of the cases correctly predicted. Further, the problem of discrepancy between the log-likelihood ratio and the goodness-of-fit has disappeared.
At least in one case, unanimity still remains a relevant determinant of control. Ceteris paribus, it increases the chance of the establishment of a management committee, compared to qualified majority, by more than 45 percent. Unfortunately, due to collinearity, this is not the case for the regulatory committee.(35)
The variable measuring conflict between the Parliament and the other Community institutions performs better, especially in case of the regulatory committee. Ceteris paribus, an increase of two rejected parliamentary amendments increases by 2 percent the probability of management control and by 14 percent the probability of regulatory control. An increase across the whole range of values that this variable takes (i.e. from zero to 20) augments by 77 and 85 percent the chance of management and regulatory control, respectively.
To conclude, McCubbins and Page’s proposition on the impact of conflict on control is strongly confirmed. Any type of operationalization we have used (procedural, Council and Parliament amendments) substantially increases the chance of some sort of ex post procedural control in the majority of models studied. Also uncertainty has a relevant impact on control, although somewhat less convincingly.(36)
So far we used the cases where there is no control as the baseline category and formulated statements in comparison with this category. We cannot say, for instance, that an increase of conflict and uncertainty leads to an increase in the stringency of control. However, since the dependent variable can be operationalized as an ordinal index it is possible to test whether there is a monotonically increasing relation between control, conflict and uncertainty.
An index of stringency of implementation control has been created according to two criteria of diminishing importance: 1) rank of political actors exercising control and 2) decision rule for referral to the Council.(37) First, the higher rank the political actor exercising control the more constrained the Commission. An implementation measure that has to be approved by the Council without the intercession of a committee of national experts becomes politically more visible. It is more likely to be put under scrutiny by the actors involved and consequently the Commission is more careful in exercising its delegated powers. In a sense, I assume that visibility decreases Commission’s autonomy in implementation. It is for this reason that I assign to the advisory committee procedure a higher value than in case of no control and to the safeguard committee procedure the highest value. For the latter case, this can also be justified by the different nature of the game. The traditional gatekeeping role played by the national experts is absent in safeguard procedures. Steunenberg (1996) has demonstrated that the discretion enjoyed by the agent is largest when a gatekeeper is involved in the game as opposed to when only veto players are present.
13
The second criterion to allocate values of the stringency index is the decision rule used in the committee to refer the measure to the Council (see second column in Figure 1). The more demanding this rule, the less likely a measure is referred to the Council, and the less likely it becomes visible and is put under strict scrutiny by the relevant political actors. For this reason, control by the management committee is less stringent than control by the regulatory committee because in the former a qualified majority is needed for referral to the Council while, in the latter, a blocking minority is sufficient for such referral. Similarly, there is no possibility of referral in the advisory committee, so very limited ex-post control is granted to other actors. The advisory committee procedure is the least strict.
Following these criteria, the index takes the value of one if a legislative act contains no implementation procedures, two if there is an advisory committee procedure, three, four and five for the management, regulatory and safeguard procedures respectively.(38) The degree of autonomy enjoyed by the Commission is inversely related to this index.
As suggested by Agresti (1990), I have employed a cumulative logit model that uses ordered dependent variables (control stringency) and forms logits of cumulative probabilities (see Appendix for more details). Figure 5 illustrates the results for the three models including 1) conflict with Council, 2) uncertainty and 3) conflict with Parliament (in the subset of cases where there is a parliamentary vote or opinion). As above, the models have been separated for problems of collinearity. Coefficients determine the cumulative probability of increasing stringency of ex post control in the J – 1 categories of the index (J is number of ordered categories).
The models perform well, with between 84 and 90 percent of cases correctly predicted, and the goodness-of-fit and likelihood ratio do not reject the null hypothesis of the perfect fit. Only the last step of the models, which measure the cumulative probability of safeguard control over the other types of control procedure, is refuted. This is probably due to the fact that safeguard procedures are predominantly used in specific circumstances, such as market disruptions and health and safety risks, that may make them independent from conflict and uncertainty effects.
Apart from this exception, the most important difference between this model and the previous ones is that unanimity loses a certain degree of statistical significance for the benefit of conflict or uncertainty variables, which are significant at 1 percent confidence level in almost all steps of the model.(39) Substantively, unanimity still tends to increase, ceteris paribus, by more than 40 percent the chance of increasing control but this is limited to lower degrees of control stringency. At step 3 of the model, this value is insignificant.
Conflict (i.e. Council amendment) and uncertainty perform statistically and substantively better that in Figure 3, especially for changes of medium-low entity.(40) Ceteris paribus, an increase of two Council amendments increases by more than 7 percent (more than 4 in step 3) and an increase of five hundred words augments by 8 percent (more than 6 in step 3) the chance of more confining control. Two rejected parliamentary amendments lead to an almost 15 percent increased probability of stricter control (more than 6 in step 3).
14
A way to interpret these results more generally could be as follows. Unanimity increases the chance of some sort of procedural control; thus it more likely determines whether there will be control. Conflict and uncertainty are more important determinants of how much control there should be, since they show a clearer monotonically increasing relation with stringency of control. McCubbins and Page’s hypotheses are hence confirmed; a restrictive committee will be preferred to a more permissive if information is highly asymmetric or if there is intense conflict between legislators.
Some students attach only an informational and coordinating role to implementation committees. Hayes-Renshaw and Wallace (1997: 182) observe that comitology ‘is a rather normal tool of the policy maker and policy implementer, namely the convening of groups through which the Commission discusses … the progress of policy implementation’. Wessels (1998) adds that comitology is the result of a process of engranage of national administrators into Community policies and stresses how the bargaining style is business-like and problem-solving.(41) This article partially confirms the thesis that implementation committees are established to limit the informational asymmetry facing Community legislators and to solve coordination problems, at least to the extent that these problems are correlated to the complexity of the issue.
However, in some cases this seems to be a prevailing view. In its report on the comitology system, the Institut für Europäische Politik observes that ‘Commission officials generally do not think that their committee significantly reduced the Commission’s freedom, and even less that it has been set up to assure the member states’ control’ (quoted from Majone, 1996: 73). In their qualitative study, Joerges and Neyer (1997: 279) add that ‘the agenda of committees is dominated by the Commission. Its room for manoeuvre is by no means substantially constrained by the shadow of majority voting which the Council included in its legislative acts.’ If committees are operationally innocuous, why is that that more conflictual policy issues are invariably linked to more restrictive control procedures? Although we cannot incorporate in the same model conflict based on amendments and uncertainty, to the extent that unanimity measures conflict we have certainly to reject the hypothesis that they perform only an informational role.
Those who see these committees in a more conflictual light as ‘an institutional response to the deep-seated tensions between the dual supranational and intergovernmentalist structure of the Community’ (Joerges and Neyer, 1997: 273) and document the strong opposition by the European Parliament (Bradley, 1997) capture more accurately, albeit partially, the institutional dynamics of the Community. The establishment of strict control procedures is also the result of substantive issue-specific conflict between Community institutions. This conclusion poses doubts on the suggestion that they have an innocuous operational impact on Commissions behavior.
15
The population includes all 1987-98 non amending secondary legislation (regulations and directives). The CELEX database and the Official Journal have been the main sources used. Unfortunately, both are slightly deficient. CELEX has some regulations whose reference cannot be found in the Official Journal. Given the legal requirements of publication, this seems to be a flaw of the database. Conversely, there is not a requirement of publication of Directives in the Official Journal, which is then incomplete in this respect. I have disregarded decisions because of their administrative and addressee-related nature, and opinions and recommendations because they are not legally binding. Directive and regulations amending decisions, protocols or conventions have been included.
Since the population shows highly skewed frequency distributions across the two key variables of policy area and legislative procedure. A simple random sample could easily under-represent a policy area or a legislative procedure. In order to decrease such sampling error without increasing the sample size, I have instead drawn a stratified random sample of 100 cases. Each stratum is characterized by a different Treaty base and legislative procedure to ensure internal homogeneity and external heterogeneity. The sample size of each stratum is proportional to the stratum population. This procedure is termed stratified random sampling with proportional allocation or constant sampling fraction. In this way, first and second order probabilities of inclusion of a case in a stratum equal simple random sampling probabilities and variance and total formulae are similar. There is no need to modify values of observations (Frosini et al., 1994: 87-8). Although only simple random sampling generates samples with independently and identically distributed cases, this proportionate stratified sampling improves representation without complicating too much the analysis (Frosini et al., 1994 : 41-45).
The cumulative logit model is a special case of the multinomial logit model that incorporates the ordering of the response variable in the construction of the logits, which are formed by cumulative probabilities. Given the need to incorporate interval-level data, I use logistic regression instead of log-linear analysis.
From Agresti (1990: 321), the cumulative logits can be defined as
Lj = log ( 
) for
j = 1, …, J –1,
where J is the number of categories of the ordinal variable (5 in our case) and pj is the probability at value x of the independent variables that a case is from the jth category. Logits of conditional probabilities are generated computing J-1 ordinary binomial regressions, re-coding cases for increasing values of the ordinal index. The likelihood-ratio and goodness-of-fit of the model has been computed by summing up the ratios of each binomial regression. This separate fitting of the model can be less efficient than simultaneous fitting, however Begg and Gray (1984) observe that inefficiency is reduced if there is a natural baseline category or if the number of cases in such category is large. The cumulative logit starts with no control as the baseline category, which fits both conditions. Thus the inefficiency of the estimators is limited.
16
Alternative methods to incorporate ordinal response variables are the continuation-ratio and adjacent-categories logit models (Agresti, 1990: 318-321). Although their results are similar, they perform slightly worse than the cumulative logit because, while cumulative logits uses all J categories, these models exclude, at a certain point, the baseline or other categories. An alternative functional form to the logit, multinomial probit, requires the assumption of a normal distribution of the cumulative density function, which in our case is not warranted (Lawrence and Arshadi, 1995; Schonhardt-Bailey, 1998).
Finally, the assumption of multinomial logit models about the independence of irrelevant alternatives(42) is appropriate in cumulative logit models for ordinal responses because the choice in the logistic regressions in our case is between more or less strictness of control, which is independent from other alternatives.
Agresti, Alan (1990) Categorical Data Analysis. New York: John Wiley & Sons.
Alter, Karen J. (1998) 'Who Are the "Masters of the Treaty"?: European Governments and the European Court of Justice', International Organization 52:121-147.
Begg, C.B. and R. Gray (1984) 'Calculation of Polytomous Logistic Regression Parameters Using Individualized Regressions', Biometrika 71:11-18.
Bradley, Kieran St Clair (1992) 'Comitology and the Law: Through a Glass, Darkly', Common Market Law Review 29:693-721.
Bradley, Kieran St C. (1997) 'The European Parliament and Comitology: On the Road to Nowhere?', European Law Journal 3:230-254.
Burley, Anne-Marie and Walter Mattli (1993) 'Europe Before the Court: A Political Theory of Legal Integration', International Organization 47:41-76.
Commission, EC (1985) Completing the Internal Market: White Paper from the Commission to the European Council. COM(85) 310 final. Brussels.
Crombez, Christophe (1996) 'Legislative Procedures in the European Community', British Journal of Political Science 26:199-228.
Demmke, Christoph, Elisabeth Eberharter et al. (1996) 'The History of Comitology'. In Shaping European Law and Policy: The Role of Committees and Comitology in the Political Process, edited by R. H. Pedler and G. F. Schaefer. Maastricht: European Institute of Public Administration.
Dietrich, Rometsch and Wolfgang Wessels eds. (1996) The European Union and Member States. Towards Institutional Fusion? Manchester: Manchester University Press.
Docksey, Christopher and Karen Williams (1994) 'The Commission and the Execution of Community Policy'. In The European Commission, edited by G. Edwards and D. Spence. Harlow, UK: Longman.
Dogan, Rhys (1997) 'Comitology: Little Procedures with Big Implications', West European Politics 20:31-60.
Edwards, Geoffrey and David Spence eds. (1994) The European Commission. Harlow: Longman.
European Parliament (1968-69) Jozeau-Marigne' Report. Document No. 115. Brussels.
Fitzmaurice, J. (1988) 'An Analysis of the European Community's Co-operation Procedure', Journal of Common Market Studies 26:389 - 400.
Frosini, Benito V., Mario Montinaro and Giovanna Nicolini (1994) Il Campionamento da Popolazioni Finite. Metodi e Applicazioni. Turin (Italy): UTET.
Garrett, Geoffrey and George Tsebelis (1996) 'An Institutional Critique of Intergovernmentalism', International Organization 50:269 - 299.
Hayes-Renshaw, Fiona and Helen Wallace (1997) The Council of Ministers. London: MacMillan.
Holmström, Bengt (1979) 'Moral Hazard and Observability', Bell Journal of Economics 4:74-91.
Hood, Christopher (1991) 'Concepts of Control over Public Bureaucracies: "Comptrol" and "Interpolable Balance"'. InThe Public Sector - Challenge for Coordination and Learning, edited by F.-X. Kaufmann. Berlin: Walter de Gruyter.
Institut für Europäische Politik (1989) 'Comitology': Characteristics, Performance, and Options. Bonn: Preliminary Final Report.
Joerges, Christian and Jürgen Neyer (1997) 'From Intergovernmental Bargaining to Deliberative Political Process: The Constitutionalisation of Comitology', European Law Journal 3:273-299.
Kiewiet, D. Roderick and Mathew D. McCubbins (1991) The Logic of Delegation. Congressional Parties and the Appropriations Process. Chicago: The University of Chicago Press.
Krehbiel, Keith (1988) 'Spatial Models of Legislative Choice', Legislative Studies Quarterly 13:259-319.
Lawrence, Edwards C. and Nasser Arshadi (1995) 'A Multinomial Logit Analysis of Problem Loan Resolution Choices in Banking', Journal of Money, Credit, and Banking 27:202-216.
Lenaerts, Koen (1991) 'Some Reflections on the Separation of Powers in the European Community', Common Market Law Review 28:11 - 35.
Majone, Giandomenico (1994) 'The Rise of the Regulatory State in Europe', West European Politics 17:77-101.
Majone, Giandomenico (1996) Regulating Europe. London: Routledge.
McCubbins, Mathew D. and Talbot Page (1987) 'A Theory of Congressional Delegation'. In Congress: Structure and Policy, edited by M. D. McCubbins and T. Sullivan. Cambridge, MA: Cambridge University Press.
McCubbins, Mathew D. and Thomas Schwartz (1984) 'Congressional Oversight Overlooked: Police Patrols versus Fire Alarms', American Journal of Political Science 28:165-179.
McFadden, Daniel L. (1984) 'Econometric Analysis of Qualitative Response Variables'. In Handbook of Econometrics, edited by Z. Griliches and M. D. Intriligator. Amsterdam: Elsevier Science Publishers.
Moe, Terry M. (1987) 'An Assessment of the Positive Theory of Congressional Dominance', Legislative Studies Quarterly 12:475-520.
Noel, Emile (1973) 'The Commission's Power of Initiative', Common Market Law Review 10:123-125.
Pedler, Robin H. and Guenther F. Schaefer eds. (1996) Shaping European Law and Policy: The Role of Committees and Comitology in the Political Process. Maastricht: European Institute of Public Administration.
Pelkmans, Jacques and L. Alan Winters (1988) Europe's Domestic Market. London: Royal Institute of International Affairs.
Pollack, Mark A. (1997) 'Delegation, Agency, and Agenda Setting in the European Community', International Organization 51:99-134.
Pratt, John W. and Richard J. Zeckhauser eds. (1985) Principals and Agents: The Structure of Business. Boston, MA: Harvard Business School Press.
Schonhardt-Bailey, Cheryl (1998) 'Parties and Interests in the 'Marriage of Iron and Rye'', British Journal of Political Science 28:291-330.
Scully, Roger M. (1997) 'Policy Influence and Participation in the European Parliament', Legislative Studies Quarterly XXII:233-252.
Steunenberg, Bernard (1996) 'Agency Discretion, Regulatory Policymaking, and Different Institutional Arrangements', Public Choice 86:309-339.
Steunenberg, Bernard, Christian Koboldt and Dieter Schmidtchen (1996a) 'Beyond Comitlogy: a Comparative Analysis of Implementation Procedures with Parliamentary Involvement', Aussenwirtshaft 52:87 - 112.
Steunenberg, Bernard, Christian Koboldt and Dieter Schmidtchen (1996b) 'Policymaking, Comitology, and the Balance of Power in the European Union', International Review of Law and Economics 16:329 - 344.
Tsebelis, George (1994) 'The Power of the European Parliament as a Conditional Agenda Setter', American Political Science Review 88:128-142.
Vos, Ellen (1997) 'The Rise of Committees', European Law Journal 3:210-229.
Wessels, Wolfgang (1997) 'An Ever Closer Fusion? A Dynamic Macropolitical View of Integration Processes', Journal of Common Market Studies 35:267-299.
Wessels, Wolfgang (1998) 'Comitology: Fusion in Action. Politico-administrative Trends in the EU System.', Journal of European Public Policy 5:209-234.
(*) Comments on previous versions of the paper from two anonymous EIoP reviewers are very much appreciated.
(1) They also suggest the interpretative key of ‘deliberative supranationalism’ that shows strong normative content and lack of analytical clarity in fleshing out causal linkages.
(2) For instance, he suggests that a realist interpretation of the existence of these committees can be gauged by a confrontational style and frequent referrals to the Council. However, also in a zero sum game if the Commission can anticipate the preferences of the pivotal actor in a committee neither a confrontational style nor frequent referrals are likely.
(3) On positive and negative power see also Krehbiel (1988).
(4) It is public in the sense that it is difficult to exclude other principals from consuming the good produced by the oversight activity (e.g. a monitoring report) (Pollack, 1997).
(5) See EC Treaty articles 85-93, 115, 130d, 48.3d respectively.
(6) Noel (1973) provides a detailed list of Treaty provisions requiring a proposal of the Commission before legislation is adopted, article 155 (EC) lists the other general instruments and article 169 (EC) gives the Commission the power to initiate infringement proceedings.
(7) See articles J.5, J.8, J.9, K.4 and K.9 (TEU).
(8) The newly approved Treaty of Amsterdam does not differ. The Treaty provides only for slight more involvement of the Commission where it is asked to submit proposals to the Council to ensure the implementation of a joint action in foreign policy and judicial cooperation (articles J.4(4) and K.6(2) Amsterdam Treaty). This however seems to be a rather limited change because proposal power is still shared with Member States and the more important common strategies are taken unanimously by the European Council leaving no leeway for the Commission to exploit conflicting preferences in the Council. The Treaty transfers also some aspects of Justice and Home Affairs into the EC jurisdiction (asylum law, migration policy, border controls and the Schengen Agreement) and amend or create new domains of Community involvement (e.g. a new Title VIa on employment and modest amendments in environment, consumer protection and public health policies, see article 2 of the Treaty of Amsterdam) but there is no change of design of the mandate. This does not apply for the Court of Justice instead, see Alter (1998: 141).
(9) If the Amsterdam Treaty is ratified, the nomination of the Presidency of the Commission will have to be approved by the Parliament (article 2.40 Amsterdam Treaty).
(10) See article 160 EC, only Commissioner Albert Borschette has been dismissed in this way because terminally ill.
(11) Other than that they must be nationals of a Member State (no more than two each) and a requirement of independence and general competence, no qualifications for candidacy are required (article 157 EC).
(12) Articles 173-5 EC specify the judicial review by the Court of Justice and article 188c EC the activity of the Court of Auditors. A Decision of the European Parliament on March 1994 has established an Ombudsman with the duty to uncover maladministration in the activities of the Community institutions. Any citizen of the Union or any natural or legal person residing in the Union may refer a complaint to the Ombudsman. This is another institution controlling Commission’s activity but similar considerations on the biased and restricted nature of control that I refer to below apply.
(13) 300 measures to be approved by 1992, see Commission (1985) and Pelkmans (1988).
(14) Dogan’s data set includes all legislation enacted from 1987 until 1995 (4601 acts), for reasons explained below I limit my data set to non-amending legislation adopted from July 1987 to September 1998 (1372 acts). The data set includes all those directives and regulations that have a Treaty base and do not modifies previous legislation, those acts that are based on a prior directive or regulation are not included because it is unclear whether they are amending. Directives and regulations amending decisions, protocols and conventions have been included if they have a Treaty base.
(15) These are areas where less than 20 new acts have been adopted, that is competition, tax provisions, economic policy, euro networks, cohesion and development.
(16) A measure of stringency of ex post control will be developed below in the analytical section, for the time being it is suffice to say that Commission’s autonomy is increasingly limited by the advisory, management, regulatory and safeguard procedures, in this order.
(17) A probable exception is social policy and customs union but I would think that this is mainly due to the difference in classification used. My classification in based on the titles of the Treaty, while Dogan’s is based on more detailed subject reference.
(18) The part of the legislative act, which is counted for the number of words, covers the text from the first article to the name of the President of the Council of Ministers included.
(19) The number of adopted amendments has been computed by comparing the final act published in the Official Journal with the Commission’s initial or, if it is the case, revised proposal. Parliamentary amendments that have been adopted by the Council have not been included; the role of the Parliament will be discussed in more details below. Council amendments that have been adopted in revised proposal have been included. Amendments can be classified into four categories: 1) spelling or grammar, 2) substantive, 3) related to policy-making functions, and 4) related to procedural requirements. Substantive amendments concern the change of technical details such as the number of tons in a tariff quota or the selection of a target area for structural funds. The third type of amendments concerns the delegation of policy-making functions to the Commission (e.g. provision of information or regulation), while the last is about the establishment of control procedures. I have disregarded the first type of amendments to compute this variable. As for the other types, they are qualitatively different but relevant to measure conflict.
(20) See below for more details on the sample.
(21) Scully (1997) suggests the following order of influence of the Parliament in the legislative procedures: consultation, co-operation, assent, co-decision.
(22) This is the so-called indicator-variable coding scheme in logistic regression. N-1 new variables are created where N is the number of categories. In our case there are two new variables, in the first one cases are coded 1 in case of unanimity and 0 otherwise while in the second variable they are coded 1 in case of co-operation or co-decision and 0 otherwise. The reference category (qualified majority) is coded 0 for both variables. Statements are made in comparison with the reference category.
(23) I need to stress that this argument transpires from Dogan’s words, it is never made explicit.
(24) The Parliament is a conditional agenda setter in co-operation and a veto player in assent and co-decision (Garrett and Tsebelis, 1996; Tsebelis, 1994). The sample does not include acts passed under assent, so the impact of this procedure is disregarded.
(25) These include number of Commission’s proposals, number of changes and of pages in Commission’s proposals, number of months passed between the initial proposal and the publication of the act in the Official Journal, number of pages of parliamentary amendments, number of specific comments and of pages in European and Social Committee and Committee of the Regions opinions. Some these variables have been dropped because theoretically less relevant (e.g. ESC opinions), other because they do not allow a clear analytical separation between conflict and uncertainty (e.g. longer time of adoption may be due either to the complexity of the measure or to the conflict between legislators), and others (e.g. number of proposals, number of pages in opinions) because of strong multicollinearity.
(26) See Appendix for details on population and sampling strategy.
(27) Since logit log-linear analysis does not incorporate interval-level data, we would need to re-code these variables into few discrete categories. Apart from the inevitable loss of information, the criteria for the allocation of cases into categories tend to be rather arbitrary.
(28) Plots and casewise listing of residuals have shown no evidence of heteroscedaticity in this and the following analyses.
(29) I have computed the Goodness-of-Fit Hosmer-Lemeshow statistic and results are similar.
(30) These are estimated probabilities using as baseline no control, that is they reflect the odds as the ratio of probability of the existence of the specific type of committee control to the probability that there will be no control. Moving from zero to 16, the odds are 61 for management, 665 for regulatory and 290 for safeguard.
(31) With an increase of two thousand, the odds are 11 for management, 81 for regulatory and 330 for safeguard.
(32) This result is confirmed also if we use a dichotomous variable for qualified majority and unanimity, leaving aside the role of the Parliament. For the management committee, the coefficients for unanimity would be 2.9491 and 3.0428 for models 1 and 2 respectively. They are both significant at 5 percent level.
(33) The interaction term between conflict and Parliament and between uncertainty and Parliament in binomial regressions is significant at 1 percent level in both models; this is not the case for unanimity. To confirm this spurious relation, Dogan (1997) observes that 50 percent of all legislation enacted under co-operation and co-decision have committee control but the Parliament still objects it.
(34) The indicator of parliamentary role has been left out because collinear with the amendment variable.
(35) The interaction term between qualified majority and rejected parliament amendments in a binomial regression is significant at 1 percent level (for this test the procedural variable has been re-coded 0 for unanimity and 1 for qualified majority). This confirms Scully’s (1997) thesis that parliamentary activism is linked to the probability of influencing final outcomes, that is the Parliament is likely to table more amendments under qualified majority rather than under unanimity.
(36) To confirm this, if we substitute uncertainty for Parliament conflict in Figure 4, this variable is statistically relevant only for the management committee.
(37) A more sophisticated index that differentiates between procedural variants using other two criteria (i.e. status quo in case of Council inaction, and timing of control – columns 4 and 3 of Figure 1, respectively) has also been used. Since it generates similar conclusions, I have kept the simpler version.
(38) In conditional logit models for ordinal responses, coding normally begins at 1 rather than zero. I have disregarded the case when there are procedures that, although termed implementation, are clearly legislative ones.
(39) The exception is parliamentary conflict in step 3, which is significant at 5 percent. I will not consider the procedural parliamentary variable in the two models in the upper part of Figure 5 because of the likely spurious relation as I have discussed above.
(40) This limit is a structural feature of logit models because estimated probabilities level off at 100 percent and differences are more difficult to detect for larger variations of the independent variables.
(41) Wessels promulgates his own idea of ‘fusion’ to comitology, which is basically a technical spillover of administrative nature which does not show the most contentious aspects of neofunctionalism, namely the transfer of loyalties.
(42) The independence of irrelevant alternatives assumes that each alternative is independent from alternatives rather than the reference category and can lead to over- and underestimation of probabilities (McFadden, 1984).