The Politics behind the Consultation of Expert Groups : An Instrument to Reduce Uncertainty or to Offset Salience ?

This paper answers the following question: Do the uncertainty and salience of issues determine whether the European Commission will use an expert group to assist with policy formulation? Using rationalist theory, three hypotheses test whether transversality, the importance of standard-setting and the salience of a policy proposal determine whether a Commission DG will ask an expert group to assist in preparing that same proposal. Data was retrieved from official documents via EUR-Lex. A binary logistic regression analysis has been conducted on a sample of 260 proposals that were drafted by DG Climate Action, DG Communications Networks, Content & Technology, DG Environment and DG Internal Market and Services. All proposals were adopted between 2010 and 2013. The empirical analysis shows that expert group involvement in policy formulation is neither negligible nor ubiquitous in terms of frequency as expert groups assisted in preparing 33.5% of the proposals. DGs were significantly more likely to consult an expert group when the proposal under preparation was more transversal in nature and/or when that proposal treated standard-setting more pronouncedly. In contrast, the salience of a proposal was shown to be insignificantly related to the presence of an expert group during policy formulation.


Introduction
"Expertise is crucial for sound policies" or at least so argues the European Commission, which regularly consults expert groups.These groups are advisory committees composed with stakeholder representatives such as member states and/or interest groups.Expert groups are-amongst other tasks-asked by Commission services or DGs to assist in the preparation and formulation of new proposals and their involvement in the policy process is especially important at this stage because it enables them to shape the content of policy (Larsson & Murk, 2007;Princen, 2011).Given the role that is attributed to expert groups in the preparation of issues, it is quite remarkable that expert groups are usually not studied by looking at issue characteristics.In spite of this, scholars agree that the European Commission uses expert groups for two reasons, namely to engage in problem-solving and to mobilise support (Larsson & Murk, 2007;Robert, 2010Robert, , 2013)).On one hand, problem-solving assumes that expert groups possess private information that is essential for the substantive quality of a proposal (Heard-Laureote, 2010).This suggests that the Commission services ask such groups to assist in the preparation of issues regarding which it experiences uncertainty.On the other hand, mobilising support assumes that experts act as representatives and that they may signal information about member state and interest group preferences regarding an issue (Haverland, 2009;Haverland & Liefferink, 2012).It is argued here that such information is especially relevant for Commission services when preparing policies on salient issues.This paper therefore asks the following question: Do the uncertainty and salience of issues determine whether the European Commission uses an expert group or not to assist with policy formulation?Until now the system of expert groups was perceived as being particularly fragmented and lacking structure (Larsson & Murk, 2007).This is especially troublesome given the overall size of a system that currently includes 358 expert groups active in policy formulation (European Commission, 2014).Some evidence suggests that the use of expert groups varies by policy area given that the Commission services which are responsible for the drafting of proposals are also in control of administering the expert groups (Douillet & de Maillard, 2010;Gornitzka & Sverdrup, 2008;Hrabanski, 2010).However, Gornitzka and Sverdrup (2008) added in a single effort that the system is as much a plethora of "issue-and policy-specific constituencies that evolve according to different logics" (p.746).Expert groups are therefore often perceived as committees that further amplify sectoral differentiation within the European Union.Hence, previous studies were far from encouraging systematic comparisons of the use of expert groups across issues.But despite the diversity and the overall size of the system, expert groups are also not omnipresent in EU policy-making as they do not assist in the preparation of each proposal.For instance, expert groups did not appear to play a role in the drafting of a major initiative such as "A Clean Air Programme for Europe" (European Commission, 2013).This is puzzling because the lack of attention to issue characteristics left scholars wondering why the European Commission was using an expert group to assist in the preparation of some policy proposals while consulting no expert group regarding others.This paper addresses that gap by testing whether issue characteristics affect the presence or absence of expert groups in policy formulation.Hereafter issue characteristics are studied along two main lines, namely that of "uncertainty" and "salience".First, and while considering that expert groups sometimes engage in problemsolving, it should be relevant to study the effect of issue uncertainty on the use of expert groups.Uncertainty points to the incapability of policy-makers to understand an issue.Second, and after taking into account that expert groups may also be used by the Commission to mobilise support, it should be relevant to study the effect of issue salience on the use of expert groups.Salience refers to the political sensitivity of an issue for member states and interest groups.
Hereafter literature on expert groups and knowledge utilisation will be introduced.Afterwards, theories on executive politics will be addressed and based on the former, three hypotheses will be formulated which link issue characteristics to the use of expert groups.Data and methods will then be discussed before the empirical results are presented.Finally, concluding remarks about the research and its implications for future work on expert groups will be presented.

Expert Groups and the Black Box of Issue Characteristics
Since the European Commission created a register for expert groups in 2005, scholars have noted an increase in the Commission's use of them (Gornitzka & Sverdrup, 2008).Their proliferation in number was one reason for a renewed interest amongst academics in the phenomenon, although issue characteristics were never explicitly addressed to explain variation and instead they remained in a figurative black-box.Pioneer contributions focused on who these experts actually were and how these groups were configured (Gornitzka & Sverdrup, 2008;Larsson & Murk, 2007).Another reason for interest resided in transparency issues regarding these groups' composition and use due to which research focused mainly on the interactions occurring inside these groups while trying to determine the logic(s) underlying their use.Academic efforts here mainly aimed at demonstrating that expertise is not used a priori in a neutral, objective or apolitical way (Robert, 2010).However, the mushrooming of expert groups by now appears to have stabilised and although controversies about transparency have remained, they no longer seem all that different from concerns addressed to other forms of committee governance in the European Union such as comitology or Council Working Groups (Brandsma, 2013;Häge, 2012).After being confronted with a general lack of transparency and data constraints in the register of expert groups, scholars continued to study expert groups from similar angles.Following a critical report by Alter-EU (2008), Gornitzka andSverdrup (2010, 2011) were again among the first to study in more detail the individual profiles of these experts, with Chalmers (2013) and Rasmussen and Gross (2014) following in their footsteps.Rimkuté and Haverland (2014) in turn explained why the European Commission actually uses expert groups, especially those composed of scientists.Taken together, there was little prospect that anyone would unpack the black box surrounding issue characteristics.
Quite recently, however, Metz (2013) explicitly addressed the matter and showed that the system of expert groups is less of a sui generis phenomenon than is often presumed.Based on semi-structured interviews and official documents, she argued that expert groups "feed into the preparatory work in multiple ways…depending on the issue context and the policy maker's corresponding demands" (p.276).While one could expect issue uncertainty and salience to play a more prominent role in such qualitative research (because data collection may pose fewer problems here), previous work was mostly constructivist in nature and tended to focus on the socialization processes occurring within these committees (Hrabanski, 2010;Robert, 2010).As such, the importance of experts to lend political authority and legitimacy to an initiative is stressed because committees are consulted to forge a European consensus (Douillet & de Maillard, 2010;Robert, 2013).
Alternatively, literature on knowledge utilisation mainly explains that policy-makers may resort to expertise for different political motives (Boswell, 2008;Radaelli, 1999).This is relevant to understanding how issue uncertainty and salience may incite Commission services to consult an expert group (Rimkuté & Haverland, 2014).The instrumental use of knowledge implies on one hand that policy-makers seek substantive advice from expert groups about issues which they do not understand adequately.Hence, whether expert groups are used according to a problem-solving logic should be linked to the level of uncertainty experienced by the Commission.Metz (2013) stated in that respect that expert groups are used mostly regarding issues that contain technical details (pp.274-275).On the other hand, the Commission may equally use an expert group to foster consensus (Douillet & de Maillard, 2010;Metz, 2013).While expert groups offer non-binding advice and experts only act as informal representatives in these groups, a consensus between national experts in favour of an issue gives the Commission a powerful argument against subsequent political opposition.Correspondingly, Metz (2013) pointed out that expert groups are also relevant for tackling "controversial" issues which, again, hints at the relevance of studying issue salience.Furthermore, knowledge is sometimes also used solely for strategic purposes in order to substantiate pre-determined policy positions (Boswell, 2008).However, this strategic use takes place rather exceptionally while the instrumental use of knowledge is considered predominant (for instance, Rimkuté & Haverland, 2014).This research therefore focuses mainly on the latter.The following section further specifies on which theoretical grounds issue uncertainty and thereafter issue salience are related to the use of expert groups by Commission services.

The Effect of Uncertainty and Salience on Expert Groups
Uncertainty is defined as the incapability of policymakers to tackle a policy problem by formulating a solution.Although uncertainty can also arise from causes rooted in a unique policy context, the present focus lies on causes that multiple proposals have in common.Two such causes, transversality and standard-setting, will be discussed hereafter.While acknowledging that other determinants of uncertainty may exist as well, transversality and standard-setting should frequently create an information disadvantage for the European Commission due to which the latter is expected to systematically seek advice from expert groups under these circumstances.An information disadvantage refers to a situation in which information is asymmetrically divided at the expense of the European Commission (Delreux, 2011, pp. 54-55).Actors that possess private information enjoy in particular an information advantage compared to other actors which are also in need of that information, and such asymmetry is in turn reflected in their bargaining position (Banks & Weingast, 1992;Calvert, 1985;Pollack, 2003, pp. 27-28).Applied with regard to EU policy formulation, this suggests that the European Commission can cope with uncertainty by consulting stakeholders such as interest groups, member states, etc. who possess private information.The more uncertain the Commission is, the more likely it is that it will seek external advice for instrumental purposes (Haas, 1992, Haverland, 2009).Admittedly, the Commission also has alternative means of expertise at its disposal for this purpose.Outsourcing consultation to a private consultancy is one such option, but a relatively expensive one.Organising workshops/seminars/etc. is another alternative, although they comprise only ad hoc meetings whereas expert groups have the advantage of meeting recurrently.This enables the latter to give advice throughout the entire process of policy formulation, which should render expert groups highly effective to tackle issue uncertainty.
Transversality points to the cross-cutting nature of policy proposals as their impact may spread across multiple policy areas (European Commission, 2009).For instance, the "Proposal for a Directive on public procurement" (European Commission, 2011) would qualify as a cross-cutting initiative because public procurement takes place in all policy areas, meaning that the initiative is of interest for most public actors as well as for the private actors that carry out tenders.As a consequence of their cross-cutting nature, transversal proposals easily exceed the competence area of the leading DG that is preparing them.Commission services are organised along functionally specialised lines (Egeberg, 2012) and so a trade-off is likely to occur between issue transversality and the problem-solving capacity of a leading service regarding that issue.Put differently, transversality is detrimental to the problemsolving capacity of individual DGs and gives way to information asymmetries.This is relevant because bureaucratic politics are at play between DGs during policy formulation (Cini, 1996).Due to their diverging policy portfolios, ideological beliefs, and other related factors, individual DGs may develop preferences for particular policy proposals which ultimately need to converge into a common position (Hartlapp, Metz, & Rauh, 2013).Although a leading DG could in principle cope with transversality by involving other DGs more closely in the preparation of an initiative, this could also weaken the bargaining position of the former towards the latter.Seeking external advice is a viable alternative as expert groups can evenly provide information about the cross-cutting nature of an issue to a leading DG.In addition, using an expert group allows the leading DG to somewhat limit the involvement of other DGs in the formulation of the initiative.The first hypothesis therefore states: H1: The more transversal an issue is, the more likely it is that a leading DG will consult an expert group.
Standard-setting describes the importance of quantifiable information (like indicators, standards, targets) for the attainment of a policy goal in a proposal.Indicators, standards or targets are referred to collectively as "quantified measures" because quantitative data is frequently essential for issues involving economic regulation (Héritier & Lehmkuhl, 2011).For instance, the "Proposal for a Regulation to define the modalities for reaching the 2020 target to reduce CO2 emissions from new passenger cars" illustrates this by name (European Commission, 2012).Quantified measures may hinder the formulation of a proposal by a Commission service in two ways.On one hand, a DG may not possess the necessary raw data (i.e.figures, numbers) to formulate a standard as the former are usually possessed by private stakeholders or by member state administrations (Héritier & Lehmkuhl, 2011).On the other hand, a DG may encounter difficulty in interpreting the adjustment costs that a standard will impose on various stakeholders (Majone, 2002).Thus, standard-setting activities should confront a leading DG recurrently with uncertainty as the DG faces an information disadvantage towards stakeholders such as business associations, but also national competent authorities who are better acquainted with the specific nature of standards and their impact on operational activities through their daily routine.In contrast to the transversality argument, a leading DG cannot resolve uncertainty caused by quantified measures through coordination with other DGs because the latter are equally prone to this information disadvantage.At this point, a leading service is expected to ask an expert group for assistance.Following a problem-solving logic, expert groups can provide information which helps the leading DG either to gather data or to estimate the policy impact of such data.That is why the second hypothesis is as follows: H2: The more an issue involves standard-setting, the more likely it is a leading DG will consult an expert group.
Salience refers to the political sensitivity of an issue (Leuffen, Malang, & Woerle, 2013).Political actors are expected to abstain from making public concessions on salient issues because salience renders proposals susceptible to heavy criticism.In anticipation of legislative decision-making, this should be worrisome for the European Commission who wants primarily to ensure the adoption of its proposals in both legislative chambers.Moreover, this should be especially problematic in the Council of Ministers where there is a tendency to strive for consensus voting (Hayes-Renshaw, Van Aken, & Wallace, 2006).Hence, when a Commission service already perceives an issue as salient at the preparatory stage or as likely to become salient later, it will intensify contacts with member states and other stakeholders in secluded meetings.This way, the Commission can build towards a consensus out of the public eye, which lowers the transaction costs of negotiation for political actors.A suitable way of doing this is by using an expert group in advance of decision-making.Constructivist theory stresses in this respect that expert groups promote diffuse reciprocity between participants (Robert, 2010).The European Commission requires that experts are familiar with European decisionmaking processes and capable of making compromises, and usually ensures that representation in the expert groups is balanced in terms of nationality.Given that meetings are restricted, expert groups are considered ideally suited for supranational deliberation and consensus-building (Hrabanski, 2010;Robert, 2010Robert, , 2013)).This in turn explains why the European Commission might again decide to consult expert groups for instrumental purposes, yet for reasons unrelated to problem-solving.When the Commission succeeds in convincing the experts to support its initiative, then their political peers will in fact have less substantive ground to keep opposing a political agreement later on.Expert groups thus have potential to facilitate decision-making and are used to mobilise support long before the onset of legislative decision-making (Larsson & Murk, 2007;Princen, 2011).The third hypothesis puts this as follows: H3: The more salient an issue is, the more likely it is that a leading DG will consult an expert group.
Elsewhere, policy nature and, in particular, the distinction between (re)distributive and regulatory politics has been posited as a powerful determinant of EU policy-making (Majone, 2002).On average, DGs who formulate regulatory policy for instance consulted more expert groups than DGs engaging in (re)distributive policy (Gornitzka & Sverdrup, 2008).No satisfactory theoretical explanation was provided for this variation, but Kassim et al. (2013) gave new impetus by further specifying this variable.While some DGs mainly focus on formulating new policies or legislation (i.e.legislative DGs), other DGs actually focus more strongly on the enforcement of existing policies (i.e. regulatory DGs) and this distinction may prove relevant when studying the variation in the use of expert groups across DGs.Given that legislative DGs mostly engage in formulating new proposals, they are expected to develop and maintain ties with diverse sets of stakeholders.Thus, legislative DGs may have a greater incentive to use expert groups in which they can meet stakeholders repeatedly.Meanwhile, regulatory DGs focus most of their attention on existing policy portfolios, due to which they may also depend more heavily on external advice when preparing new policies.
In addition, others approached uncertainty in terms of legal complexity and studied the concept in relation to adopted legislation (for example, see Klüver, 2013;Reh, Héritier, Koop, & Bressanelli, 2013).They argue that uncertainty is reflected in the length of legal acts, their number of recitals, their number of legislative articles, etc. because 'complex' legislation requires 'detailed elaboration'.In line with this legalist focus, one can alternatively argue that the drafting of legislative or legally-binding proposals-on average-creates higher transaction costs for a leading service than the drafting of proposals that are non-legislative or not legally-binding.Henceforth, a leading service can be expected to consult expert groups, especially when drafting legislative proposals.

Data and Method
Each policy proposal is considered an individual case and cases were identified through EUR-Lex. 1 EUR-Lex is an online database which gathers public documents is-  The dependent variable use of expert group is dichotomous because the European Commission either uses an expert group to assist in policy formulation (= 0) or not (= 1).The information gathered in the register of expert groups is inadequate for cross-case comparisons, but Commission documents proved a valuable alternative source for this information.Most importantly, the COM-documents representing an issue usually contain a section in which a leading DG explains how it conducted consultation during the course of policy formulation.The relevant section in a COM-document is usually titled "Results of consultations with the interested parties and impact assessments" and one way to consult is of course by using an expert group.Occasionally, such information is provided elsewhere in the preamble of the proposal instead.Otherwise, one can look for involvement of expert groups in Impact Assessment reports and Roadmap documents which occasionally accompany COM-documents.In these documents a leading DG needs to justify how it took "Consultation and expertise" into account or in which way "stakeholders and experts have been consulted".
When the European Commission was reported to have consulted a group/committee/etc. in any of these sources, the register of expert groups was checked to see whether this alleged expert group actually corresponded with a registered one.The dependent variable was then coded "1" while it was coded "0" for all other instances.However, the Commission may also announce in a proposal that an expert group has been established to assist with the formulation of related initiatives thereafter.In such a case the dependent variable is also coded as "0" because the expert group did not yet play any actual role in preparation of the initiative in question.
From 2010 onwards a total of 99 expert groups supposedly assisted one of four selected Commission services in policy formulation.Table 1 depicts the proportion of expert groups used by each DG and this approximates the proportion of issues formulated by the DGs quite well.However, the register of expert groups seems to overstate the involvement of expert groups as the latter were only used in 33.5 % of the cases in the sample (see Table 1).Thus, it is possible that expert groups were registered to assist -among others -in policy formulation while they were only consulted during the sample period for other purposes.Nonetheless some caveats need to be considered.On one hand, some expert groups assisted with preparing more than one proposal.In contrast, the Commission sometimes also asked multiple expert groups for advice about the same initiative.This cannot be inferred from the aggregated data in Table 1.
Independent variables are measured by multiple indicators which are based mostly on procedural information (see Table 2).Multiple-indicator measurement is used because measurement validity benefits from triangulation.Besides, large-N analysis is facilitated by decomposing thick concepts such as uncertainty and salience into several indicators for measurement because the latter focus on distinct properties of the original concept and therefore capture the broader meaning of that concept (Coppedge, 1999).
Firstly, transversality is a continuous measure whose operationalisation is based on two procedural indicators.On one hand, it was noted how many Commission services took part in an inter-service consultation organised by the leading service.During this internal meeting, the leading DG reports its progress on the drafting of an initiative to all other concerned DGs (Hartlapp et al., 2013).Relevant information was obtained from the Secretariat-General through personal correspondence.On the other hand, the number of European Parliament committees that formulated an opinion about the adopted initiative was also measured.Relevant information was retrieved via the Legislative Observatory.Despite the main focus on the European Commission, there is no reason why the number of parliamentary committees should not vary along with issue transversality in a similar direction as the number of Commission services would.In addition, this also allows for inferences about cases for which information on inter-service consultations is missing (see Table 2).It is argued that the more transversal an issue is, the more DGs/committees will show interest in policy formulation and decision-making.Next, the scores on both indicators were standardised in order to make them comparable and to combine them into a single measure.
Secondly, standard-setting is a continuous measure whose operationalisation came about in three steps.COM-documents were first searched for the following terms: "standards", "standardi", "indicator" and "target".The European Commission is particularly argued to experience structural information deficits concerning proposals that involve indicator-, standard-or target-setting.Next, the resulting search hits were all summed up and weighted by the length of the text document in which the word search took place.
Third, salience is measured through two indicators which capture the amount of attention that member states or private stakeholders pay to an issue (Warntjen, 2012).On one hand, it is measured whether the Commission already presented information about the initiative to the member states at a formal Council meeting before policy formulation was concluded (No = 0, Yes = 1).An initiative should, however, be mentioned in the meeting agenda under the heading "Any other business" as the other headings concern initiatives which have already been adopted by the Commission.On the other hand, the number of contributions in response to a public/online consultation is noted because it reflects the amount of attention that an initiative attracted from stakeholders (Klüver, 2013).When the Commission did not organise a public consultation regarding an initiative, the latter was coded as "0".Missing values were attributed when the number of responses to a public consultation was unknown.
Lastly, two control variables are included in the analysis.Commission service is a categorical variable and indicates which Commission service drafted the policy initiative (DG Connect = 0, DG Clima = 1, DG Environment = 2 and DG Markt = 3).Legal act is a dichotomous variable that indicates whether a case concerns a non-legislative (= 0) or a legislative proposal (= 1).The former applies to 40% and the latter to 60% of cases.
The data was analysed using binary logistic regression analysis because the dependent variable is dichotomous.A Hosmer-Lemeshow test confirmed that logistic regression is an appropriate technique to analyse the sampling data (Field, 2013).The effects of the independent variables on the dependent variable are calculated using maximum likelihood estimators.Put differently, the study examined whether issue transversality, the importance of standard-setting and the salience related to particular issues make it less/more likely for the Commission to use an expert group to assist in policy formulation.

Empirical Analysis
Table 3 shows the results of three regression models.
Model 1 is a baseline model and only contains the variables relating to issue uncertainty and salience, whereas Model 2 also contains the control variables.Model 3 presents a full model which includes an interaction term of transversality and standard-setting as well.The individual effects of these independent variables on the use of expert groups are demonstrated by coefficients.The latter are in fact odds ratios which represent the change in odds occurring when a particular category of an independent variable is met.When the odds ratio has a value smaller than 1, this means that the Commission is less likely to use an expert group under the given circumstances.In return, the likelihood that the Commission will consult an expert group increases when an odds ratio is larger than 1.In the sample of 260 cases, 33 cases had missing values and 3 cases exerted a disproportionate influence on the model.These were excluded from the analysis which was eventually performed on 224 issues.A comparison between the three models demonstrates that each model scored statistically significantly, but the full model is capable of explaining 23.6% of total variance while the other two models only explain 12.6% and 18.5% as shown by their respective R-square measures.
-2LL expresses the amount of variance that a model leaves unexplained.Again, a comparison shows that the unexplained variance decreased most for the third model.Hence, the overall model fit improved after adding the control variables and did so again when the interaction term was added.Hereafter the results of Model 3 will be discussed in more depth.Firstly, transversality is related in a statistically significant way to the use of expert groups.The direction of the effect is positive, as a leading service is more than twice as likely to consult an expert group on proposals that are more transversal.Thus, the evidence strongly supports H1 which attributes this to the fact that leading DGs have a high level of discretion regarding the way in which they use expert groups.This nuances other recent findings about bureaucratic politics within the European Commission.Rivalry between different Commission services used to be considered detrimental for the coherence of policy formulated by the Commission because individual DGs tried to further their own policy goals without taking the policy goals of other DGs into account (Cini, 1996).However, more recent work has suggested that this situation has improved considerably as the Barroso presidency promoted horizontal procedures to ensure consistency in policy formulation and also strengthened the role of the Secretariat-General in overseeing such coordination (Kassim et al., 2013).Internal consultations and procedures should therefore provide the Secretariat-General and other concerned services with ample op-portunities to ensure that a proposal does not work counterproductively relative to proposals prepared by other services.This suggests that the European Commission has become more effective in coping with forms of uncertainty that arise from transversality than it used to be because the rationale that underpins these internal consultations is one emphasising the cross-cutting nature of proposals.Yet, the data shows that leading DGs are also more likely to meet expert groups regarding transversal issues, presumably because the latter are helpful in establishing or maintaining their privileged position in the process of policy formulation.This suggests that expert groups weaken horizontal coordination within the Commission although concerned Commission services occasionally attend expert meetings as well.Further research should test whether expert groups actually help a leading DG to keep proposals under preparation below the radar of other services or whether this relation is in fact a less contentious one.
Secondly, standard-setting also has a statistically significant effect on the use of expert groups.The odds ratio is larger than 1, meaning that the assistance of an expert group in the drafting of a proposal becomes more likely along with the importance of standardsetting regarding that issue.This confirms the argument raised by H2.Of course, it is generally accepted that standard-setting may create an information disadvantage for political actors such as Commission services, but empirical confirmation of the fact that Commission services use expert groups in a systematic way to address such deficits is completely new.Previously, it was suggested that an expert group could assist in collecting relevant data or that it could provide advice about the impact of proposed measures.Based on the results depicted in Table 3, it is not possible to infer directly which reason(s) hold(s) true the most.Yet, it seems rather unlikely that the Commission would lack relevant data so frequently because the raison d'être of many regulatory agencies is exactly to gather, analyse and make available such data for the Commission.Furthermore, so-called European Standardisation Organisations are also active in creating common standards.For these reasons it seems more plausible that the Commission asks expert groups for feedback regarding estimated adjustment costs (whether the latter are proportional in nature, not discriminatory, etc.).Most suspicion surrounding the participation of private stakeholders in expert groups seems grounded in this context, where the risk for double-hatted experts is of course always imminent.
However, the effects of transversality and standardsetting should not be considered in isolation from one another as their interaction term also scored significantly.Beforehand, one would expect their combined presence to render expert involvement more likely because a Commission service should face a severe information deficit when drafting a cross-cutting issue which involves standard-setting on top.Yet, the coefficient of the interaction term is smaller than 1, which means that it actually mitigates the individual effects of transversality and standard-setting on the use of an expert group.Despite this mitigating effect, the likelihood of using an expert group does still increase for cross-cutting issues which also involve standard-setting in comparison with issues that are only transversal or only involve standard-setting.Rather, the coefficient of the interaction term indicates that the odds do not in-crease exponentially.It is not clear-cut what this means, but it seems improbable that a single expert group can solve an information deficit which is rooted in different causes.As noted in the theory section, transversality and standard-setting require different sets of expertise.Multiple expert groups might be consulted to address these problems separately, but then again this would also require additional coordination between these groups.In this sense expert groups should neither be considered as a one-size-fits-all solution.Thirdly, salience scored insignificantly, thereby proving H3 to be incorrect.It could already be observed that issue salience alone could not account for the presence of all expert groups in policy formulation as the majority of proposals are not salient.This was also the case for other research working with a full sample (Reh et al., 2013).By way of illustration, only 17.3% of all issues were discussed during a formal Council meeting, while the number of responses to public consultations clearly followed a right-skewed distribution.The reason why H3 was not confirmed is probably because the Commission perceives salience in a more ambiguous way than was assumed.It was argued before that salience impedes decision-making which should give Commission services an incentive to use an expert groups for consensus-building.However, the Commission does not necessarily dislike salience.For instance, a Commission service may equally try to gain more attention for a proposal when public opinion seems to favour its policy position.In doing so, the DG then pressures other political actors to concede with its proposal.Evidently, referring a proposal to the secluded stage of expert groups would run counter to such ambition and hence, this could explain why H3 was not confirmed.One could object that even under such circumstances a Commission service might consult an expert group for a more strategic purpose; experts can lend political legitimacy to the Commission by backing up a proposal with scientific argumentation (which makes it harder for opponents to discharge a proposal as being biased).Yet there is no reason why Commission services should use expert groups systematically in this strategic sense.In fact, previous research pointed out that this strategic use of expert groups remains the exception rather than the rule (Rimkuté & Haverland, 2014).
Lastly, the control variables presented a mixed providence in predicting the outcome variable.Regarding "Commission service", the distribution of expert groups differs across DGs, as was already shown in Table 1.The differences were quite subtle between DG Clima, DG Connect and DG Markt which is again reflected in the coefficients in Table 3.However, DG Environment really stood out as a low user, which is why its odds ratio scores below 1.The relationship further appears statistically insignificant in all models meaning that variation in use of expert groups-although notable in the first instance at the level of DGs-is better explained by the issue characteristics.Regarding "legal acts", expert groups are two and a half times more likely to be consulted about legislative proposals than regarding non-legislative proposals and this effect is significant.This makes sense in that the latter involves, for example, Commission Communications such as "A Clean Air Programme for Europe" which announce a future strategy rather than proposing detailed policy measures.

Conclusion
This paper examined whether issue characteristics affected the use of expert groups by the European Commission.An expert group was found to be present in 33.5% of policy proposals.When reverting to the initial puzzle, the analysis showed that Commission services use expert groups as an instrument to reduce uncertainty and not as a means to offset salience.These findings also speak to other research on expert groups because even though the system of expert groups constitutes a diverse patchwork at first sight, and thereby reflects to a large extent the sectoral differentiation within the European Union, this does not imply that expert groups have nothing in common across the borders of the respective policy niches or policy areas in which they work.For one thing, the results showed that issue characteristics have explanatory potential when studying expert groups.As long as more thorough assessments of expert groups through large-N studies are obstructed by data constraints, it should not be taken for granted that the expert group system itself lacks consistency.Therefore, further research could look for determinants of uncertainty other than transversality and standard-setting.In this view research on expert groups may benefit from looking at research that studies issue characteristics in relation to decision-making processes in other executive institutions.
Finally, Commission expert groups are frequently contested for their secrecy, although such criticism is usually grounded in more general concerns about secluded decision-making.Committees with restricted access, such as expert groups, are considered to increase the efficiency of executive decision-making.However, these efficiency gains also incur penalties regarding the political legitimacy that executive institutions such as the European Commission enjoy.This is so because the secluded nature of expert groups conceals how and under whose instigation politically relevant decisions came to life.Similarly to what has been seen before in the context of the comitology system, the European Commission is nowadays repeatedly being asked to make its expert group system more transparent.This has resulted in some minor concessions in previous years, but major improvements seem rather unlikely in light of the inherent trade-off that would occur between the efficiency and legitimacy of decision-making in expert groups.This research has highlighted some circumstances in which the European Commission is likely to use expert groups and, in doing so, these findings may enable scholars to make more finely-tuned normative assessments about whether such expert involvement can sometimes be justified, rather than contesting a priori that it is not.

1
EUR-Lex is accessible via http://eur-lex.europa.eusued by EU institutions.In 2014, the database merged with PRE-Lex which was formerly known to document legislative drafting.As such, EUR-Lex now compiles information about proposals (termed preparatory acts in the database) and decision-making procedures.The research sample includes nearly all proposals which were drafted by one of four selected Commission services and subsequently adopted by the College of Commissioners in the period between 2010 and 2013.The Commission services in question are DG Climate Action (DG Clima), DG Communications Networks, Content & Technology (DG Connect), DG Environment (idem) and DG Internal Market and Services (DG Markt).The sample was chosen to be diverse in terms of policy nature(Kassim et al., 2013).DG Climate Action and DG Environment are considered 'legislative DGs' whereas DG Markt is a 'regulatory DG'.Besides being a 'regulatory DG', DG Connect also engages in redistributive activities due to which it can be considered representative of those DGs that administer more hybrid policy domains.This sample allows us to test the policy nature variable empirically and to generalise causal inferences for other policy-making DGs.More specifically, the sample includes legislative proposals for regulations, directives, decisions and non-legislative proposals such as Commission communications, green papers, recommendations and white papers.No Commission opinions met the sampling criteria.Commission reports fell outside the scope of this research as they generally involve evaluation or implementation rather than the formulation of policy.This totalled to 260 cases as is shown in Table1.Most cases in the sample were prepared by DG Markt followed by DG Environment, DG Connect and finally DG Clima.

Table 3 .
Logistic regression models of "use of expert group".