Legal Instrument Choice in the European Union

Regulations and directives are the central legal instruments used by the EU. In some instances, the Commission is not legally required to choose a specific legal instrument, but can make this decision autonomously. However, we know surprisingly little about the factors that influence this decision. Based on an original dataset of all directives and regulations proposed by the European Commission in ordinary legislative procedures between 2009 and 2018, we find that the choice of a legal instrument is strongly determined by prior policy decisions and varies systematically across policy areas depending on the extent to which they have traditionally been addressed under the co-decision procedure. In addition, we find that the Commission’s use of regulations increases under conditions of increased euroscepticism, indicating that instead of granting dissenting member states more room to manoeuvre, the Commission prefers to keep them on a short leash.


Introduction
EU secondary legislation is mainly based on two types of legal instruments: directives and regulations. While the provisions set out in both instruments are applicable in all member states, the key difference between a directive and a regulation is that the former is 'binding, as to the result to be achieved, (…), but shall leave to the national authorities the choice of form and methods', whereas the latter 'shall be binding in its entirety and directly applicable in all Member States' (Treaty on the Functioning of the European Union [TFEU], Art. 288). This implies that member states enjoy leeway when transposing directives into national law (Zhelyazkova et al., 2018), while the national level is essentially skipped when it comes to regulations.
In some instances, the TFEU prescribes the use of a certain instrument and if it does not, the Commission is bound by the principles of subsidiarity and proportionality when deciding whether to issue a regulation or a directive (Art. 296, TFEU). In reality, however, the Commission often makes an autonomous decision over what it considers to be in line with these principles when issuing its policy proposals. As a result, there is a range of policy topics in which the Commission enjoys discretion over the choice of the legal instrument, and by implication, the amount of discretion it grants to member states to decide on the ways and means to reach a certain policy goal.
This discretion can be best illustrated by a direct comparison of the justifications for the choice between two different, but quite similar legal acts. In the explanatory memorandum of the proposal for the 'Directive on Copyright in the Digital Single Market', the Commission justifies its choice of directive by saying that '[i]t also leaves (…) margin of maneuver for Member States while ensuring that the objective of a functioning internal market is met' (European Commission 2016a, p. 6). In the context of the 'General Data Protection Regulation' (European Commission 2012), by contrast, the Commission argued that using a regulation is the right choice given that it 'reduce[s] legal fragmentation and provide[s] greater legal certainty by introducing a harmonized set of core rules (…) and contributing to the functioning of the Internal Market' (p. 5). According to the TFEU, both proposals could have been either a directive or a regulation. Accordingly, it seems that the Commission is, at first sight, somewhat arbitrary in choosing one legal instrument over the other. After all, would it not be also appropriate to opt for a regulation on copyright issues in order to avoid 'legal fragmentation'? Or would it not be appropriate to grant the member states some 'margin of maneuver' when implementing data protection rules?
Whether the Commission decides to issue a regulation or a directive on a given issue has substantial consequences for the governments of member states. The Christian Democratic Union (CDU), for instance, which currently forms the German federal government together with the Social Democratic Party, announced that it voted in favour of the controversially debated revision to the Copyright Directive only because it could still make use of the possibility of adjusting crucial aspects of the legislation during transposition at the national level (CDU, 2019). Despite these considerable impacts, we still lack a clear understanding of the reasons that drive the Commission to pick one legal instrument over the other when the TFEU leaves it to the Commission to decide which instrument to use.
In this contribution, we theorize and test several hypotheses on the drivers and determinants of legal instrument choice in the EU. Under which conditions does the Commission prefer to issue regulations and thereby keep member states on a short leash? And under which conditions is it willing to grant the member states discretion at the transposition stage by issuing a directive? Our analysis builds on a novel dataset covering all regulations and directives proposed by the European Commission under the ordinary legislative procedure since the Lisbon Treaty entered into force on 1 December 2009. We excluded special legislative procedures from the scope of our analysis to hold procedural scope conditions constant. Our dataset systematically distinguishes between legal acts for which the Commission can decide which legal instrument to use and those for which it cannot.
We find that the Commission's legal instrument choice depends on several factors. First, and most importantly, the legacy of prior decisions plays a crucial role. Once the Commission has opted for a given legal instrument, it hardly ever deviates from this choice in future revisions, even in cases in which the entire law is repealed and replaced. Second, we find substantial variation across policy areas. In one set of policy topics the Commission is more often inclined to grant member states discretion (for instance in the area of transport and environmental protection) and in others it rarely or even never does so (for instance in the areas of civil liberties, agriculture and fisheries) even though it could, based on the TFEU. The use of directives is particularly pronounced in areas in which the European Parliament (EP) had traditionally been involved as a co-legislator. Finally, we also find significant variation in legal instrument choice over time. In particular, the increasing euroscepticism in the Council was accompanied by an overall drop in directives over the period of our investigation. Juncker's approach to streamline the Commission's agenda to regain legitimacy might thus be considered a somewhat ambivalent strategy. While previous studies show that the Commission has reduced the overall volume of its legislative initiatives, we found that it did so by granting less discretion to member states. The remainder of this article is structured as follows. In section 1 we provide an overview of the existing literature. We show that political science scholars have focused their analytical efforts mainly on directives, while the efforts of legal scholars have often remained descriptive in nature. In section 2 we introduce several theoretical propositions on the factors that influence the Commission's decision on the legal instrument. In section 3 we introduce our data and methods and in section 4 we provide an empirical test of our theoretical propositions. Section 5 presents our conclusion.

I. Why Does the Choice of Legal Instruments Matter?
The choice of legal instruments in the EU has raised hardly any interest among scholars of European integration, at least among those with a political science background. Few studies analyse the impact of using different legal instruments, most prominently those dealing with explaining the duration of the decision-making process. The main finding in this strand of literature is that directives usually take longer to be adopted than regulations and decisions (for example, Hurka and Haag, 2019;Klüver and Sagarzazu, 2013;Rasmussen and Toshkov, 2013). To a large extent, scholars of European integration have focused on analysing directives only. As a result, almost the entire body of EU literature on policy implementation and compliance focuses on EU directives (Falkner et al., 2005), while regulations have received hardly any attention (Treib, 2014, p. 16). This is mainly due to two reasons. First, the very fact that directives must be transposed into national legislation has provided researchers with an easy way to approximate, quantify and compare countries' implementation performance by simply referring to the duration and adequacy of the transposition process (Angelova et al., 2012). This focus, however, has largely neglected the fact that transposition is only one part of the implementation process, followed by the application and enforcement of legal measures (Versluis, 2007). Second, directives are generally expected to raise significant political conflict, whereas regulations have the reputation of being purely technical and often largely uncontroversial pieces of legislation. However, this view is wrong. When we compare how often regulations and directives are put on the Council's B-agendathat is, where particularly controversial policy proposals are listedwe find little meaningful difference between regulations and directives (König, 2008, pp. 149-154). This also holds true if we look at more recent time periods. Of the 331 policy issues that have been identified as controversial in the Decision-Making in the EU II dataset (DEUII) between 1996 and 2008, 171 were found in regulations (52%), 136 in directives (41%) and 24 in decisions (7%) (authors' calculations based on data in Thomson et al., 2012). These figures clearly suggest that the view that regulations are merely technical and uncontroversial is mistaken. The 'European Regulation on Registration, Evaluation, Authorization and Restriction of Chemicals (REACH)', for instance, was issued as a regulation, but stands out as one of the most controversial law-making processes in EU history. In sum, the aspects presented indicate that a sole analytical focus on directives is misguided. Both legal instruments are crucial elements of the EU's legal framework, while each has its own strengths and limitations (Siedentopf and Ziller, 1988). While the political science literature has been largely blind to the issue of legal instrument choice due to its strong focus on EU directives, legal scholars have, at least in part, addressed the topic (Bast, 2003;Engel, 2018;Hofmann, 2009). For instance, von Bogdandy et al. (2004) assessed the proliferation of different legal instruments and studied which EU institutions were key in promoting the use of Legal Instrument Choice in the European Union specific legal instruments prior to the Treaty of Nice. Yet, while these studies are highly informative in descriptive terms, they do not provide an answer to the analytical question of why the Commission chooses a given legal instrument under different circumstances. It is this latter aspect that we intend to shed light on.

II. Theory
In this section, we develop three hypotheses on the factors that might influence the Commission's decision on the legal instrument to use. We expect that the Commission's choice is determined jointly by remote structural factors as well as by proximate strategic considerations. Here, we focus exclusively on proposals in which the Commission can make the decision either to opt for a directive or a regulation. This excludes all situations in which the legal instrument is already predefined by the TFEU.

Policy Legacy and Legal Instrument Choice
The Commission typically does not start from scratch when proposing legislation. In most of the cases in which the Commission decides to take action, it amends existing legislation. Here, we can expect that once the Commission has opted for a given legal instrument, it will not undo its decision in future revisionseven when a given legal act is completely repealed and replaced. There are several reasons for this.
First, the initial choice of a certain instrument influences the expectations of the member states on the future trajectory of EU law. Regulations create legal certainty and establish a level playing field across the EU through the complete harmonization of processes and standards. Repealing regulations and replacing them with directives would thus decrease legal certainty and hence increase monitoring costs for the Commission, as it would then be required to check member states' compliance. The Commission will thus replace regulations with directives only in the rare instances in which it considers that the loss of legal certainty and the increasing costs of monitoring compliance are lower than the benefits of allowing member states to adopt their own national laws.
Similarly, the replacement of directives through regulations also appears unattractive. Here, it is not the Commission but member states that might oppose a change of instruments. The Commission needs the approval of the Council in order to steer its proposals through the ordinary legislative procedure, which implies that it needs to factor in the costs that member states pay when formulating its policy proposals. Switching from directives to regulations requires member states to accept (completely) uniform legal standards and thus to set aside any existing national provisions. Changes in legal instruments should thus be limited to the rare cases in which the existing legal framework has clear and undeniable deficiencies and shortcomings. In this context, the so-called 'dieselgate' scandal is a perfect example. Here, the Commission (backed by the Council) promoted a shift from a directive to a regulation due to the existence of unequal implementation practices across the EU. These instances, however, are rare exceptions rather than the rule.
Second, beyond these political considerations, legal imponderables are involved when switching between different legal instruments. While it is generally possible to replace a regulation with a directive, and vice versa, the legal services of the EU institutions do not recommend this practice. They state that 'it is preferable for an amending act to be of the Steffen Hurka and Yves Steinebach same type as the act to be amended' (see rule 18.7 in European Commission, 2016b, p. 60). Thus, while switching between legal instruments when amending EU law is possible, it is broadly considered bad practice, as it comes with the danger of creating legal confusion and uncertainty. All this underlines the importance of the choice of the legal instrument, as it demonstrates that the Commission makes far-reaching decisions when proposing the first piece of legislation on a certain matter. Once a directive or a regulation has been published in the Official Journal, the Commission will find it very difficult to expand or constrain the member states' room for manoeuvre later on, both for political and for legal reasons. The respective hypothesis thus reads as follows: H1 Legal instrument choice is determined by the choices the Commission has made in the past.

EP Involvement and Legal Instrument Choice
The previous hypothesis posits that the Commission is often greatly constrained by decisions it has made in the past. Yet this does not explain why the Commission chooses one legal instrument over another in the first place. To explain today's legal instrument choice, we thus need to theorize how things differed in the past and how these aspects might still be reflected in recent policy decisions.
With the coming into force of the Lisbon Treaty in 2009, the involvement of the EP as a veto player has become standard practice under the ordinary legislative procedure, formerly known as co-decision. Yet, before 2009 co-decisions were much less widespread than today and the consultation procedure, in which the EP does not enjoy the right to veto legislation, was still broadly applied. While both procedures still exist today, consultation is nowadays considered a special legislative procedure whose use has declined considerably. Yet we expect that the traditionally enhanced role of the EP in certain policy areas should influence the Commission's choice of legal instruments today.
On the one hand, we could expect the Commission to be more inclined to use regulations in policy areas in which the EP has a history of involvement as a co-legislator. In general, the EP is known for its support of strong European integration and for its preference for member states to have less discretion due to its lack of independent monitoring capacities in such cases (Franchino, 2007). On the other hand, the involvement of the EP in the co-decision procedure has typically drawn more public and political attention to a dossier and thus increased the degree of politicization in both the EP and the Council (Häge, 2011). Under these circumstances, the greater flexibility of a directive could facilitate inter-and intra-institutional negotiations and might therefore be considered to be a more efficient instrument than the rigid regulation. Thus, under the co-decision rule, a strategically oriented Commission might have a tendency to issue its policy proposals as directives rather than as regulations to solve inter-institutional conflicts during the negotiation phase. Although the existing literature provides logical foundations for both expectations outlined above, we consider the second view somewhat more convincing for two reasons. First, the general assumption that the EP has a strong preference for deeper integration underlying the first view has become increasingly difficult to uphold, given the surge of the eurosceptic vote in the two most recent elections. Second, recent research Legal Instrument Choice in the European Union suggests that the involvement of the EP has instead led to a decrease in the legislative productivity of the Commission (Rauh, 2018), which suggests that the EP is not necessarily always only a willing ally of the Commission, but is also a potential veto player (Tsebelis, 2000).
Given the lasting effect of the Commission's past decisions hypothesized above, we assume that the different number of institutions involved in the legislative process prior to the Lisbon Treaty are still reflected in present cross-sectoral variance in the choice of a legal instrument. More precisely, we expect that the Commission is more likely to resort to the use of directives in areas that have long been addressed under the co-decision procedure. The respective hypothesis reads as follows: H2 The Commission is more likely to resort to the use of directives in policy areas in which the EP had traditionally been involved as a co-legislator under the co-decision procedure.

Euroscepticism and Legal Instrument Choice
Under the impression of the Euro and migration crises and the success of eurosceptic parties in national elections, many governments have become more sceptical of the need for deeper European integration. While the Commission has traditionally countered anti-EU tendencies in the Council by crafting coalitions among the progressive member states (Häge, 2013;Kaeding and Selck, 2005), this strategy is becoming more and more difficult in an increasingly eurosceptic environment. The recent literature thus predicts a shift 'from [a] permissive consensus to [a] constraining dissensus' (Hooghe and Marks, 2009) that represents serious obstacles to supranational policymaking. From a theoretical perspective, rising euroscepticism could lead either to more regulations or to more directives. On the one hand, if the Commission is concerned about an increasing danger of non-compliance due to rising euroscepticism it should prefer the use of regulations, which do not need to be transposed and take direct effect. As non-compliance increases when public support for European integration wanes (Fjelstul and Carrubba, 2018), the Commission should face incentives to respond to rising euroscepticism by issuing more regulations. On the other hand, one could also make the case that the Commission is not necessarily concerned solely with potential non-compliance but also with the perceived legitimacy of its policy proposals. Among eurosceptic governments, the legitimacy of a directive is naturally higher, because it leaves the member states more room to manoeuvre. This is nicely exemplified by the Austrian government's recent explicit demand to the Commission that 'directives should take precedence over regulations in order to give national and regional parliaments the chance to be involved in the legislation under discussion' (Lopatka, 2019, p. 34). Accordingly, a valid alternative strategy of the Commission to secure support for their policy proposals might be to resort to the use of directives to give member states' governments the chance to take account of their domestic regulatory contexts and their citizens' needs. Given these divergent theoretical expectations, we formulate the following competing hypotheses on the role of euroscepticism for the choice of legal instruments: H3a Due to concerns over potential non-compliance the Commission should be more likely to adopt regulations when euroscepticism increases in the Council.
H3b Due to concerns over the perceived legitimacy of its policy proposals the Commission should be more likely to adopt directives when euroscepticism increases in the Council.

III. Data and Methods
Our analysis focuses on secondary legislation proposed by the Commission under the ordinary legislative procedure since the coming into force of the Lisbon Treaty on 1 December 2009. Thus, unlike other studies, which analysed vertical shifts of legislative activity from secondary to tertiary acts (Junge et al., 2015), we look at horizontal shifts between different legal instruments. Our dataset contains all Commission proposals issued in the time period under scrutiny and listed in the EUR-Lex database. For each observation, our dataset indicates (1) the date of the proposal; (2) the legal instrument used; (3) the legal basis as specified in the TFEU; (4) and whether the respective legislation is completely new or repeals and amends existing acts. Moreover, we identified (5) the respective policy area by referring to the lead committee in the EP dealing with the proposal. In total, our dataset contains the 886 directives and regulations that the Commission proposed during our investigation period. We exclude decisions for two main reasons: first, unlike directives and regulations, decisions can be (and often are) addressed to individual actors and are accordingly not always relevant to all member states, limiting their legal comparability. Second, EUR-Lex is incomplete with regard to decisions with addresses (Blom-Hansen, 2019, pp. 701-3). In a second step, we manually coded whether the Commission had discretion in choosing the legal instrument. More precisely, we went through every TFEU article referenced in the proposals and checked whether the proposal's legal basis granted the Commission discretion in choosing the legal instrument. To do so, we consulted the TFEU as well as numerous legal commentaries and analyses. One criticism that could be raised against our approach is that the Commission might pick a proposal's legal basis strategically in order to increase its leverage when choosing the legal instrument. We argue, however, that the Commission's ability to do so is severely limited for two main reasons. First, the European Court of Justice ruled in 1987 that the 'choice of the legal basis for a measure may not depend simply on an institution's conviction as to the objective pursued but must be based on objective factors which are amenable to judicial review' (European Court of Justice, 1987, paragraph 11). Second, the legal bases that prescribe the use of a certain instrument are often very precise 1 and it would appear to be difficult for the Commission to switch to or from these legal bases when developing a policy proposal (whether the Commission switches between different legal bases that allow for discretion is not relevant in our analytical context). And in fact, the TFEU grants the Commission a lot of discretion when choosing a specific legal instrument. In important policy areas such as transport, maritime and aviation policy the Commission is free to choose which legal instrument to select. Table SI in the Annex summarizes the respective provisions laid down in the TFEU.
These legal scope conditions have important ramifications for our research strategy. In order to test our theoretical expectations, we relied only on the cases in which the Commission actually had a choice. In 35 instances, the Commission proposal related to several legal bases in the TFEU simultaneously, which contradicted each other in terms of whether the Commission had discretion in choosing the legal instrument. We decided to drop these cases from the analysis, as we considered it impossible to figure out in retrospect which legal basis the Commission considered decisive for its final choice. In 13 instances, we could not identify the lead committee in the EP dealing with the proposal based on the information available in EUR-Lex and we dropped these cases as well. Out of the remaining 838 directives and regulations the Commission proposed during our investigation period, the respective legal basis in the TFEU tied the Commission's hands in 446 cases. This corresponds to 53 per cent of all cases in our dataset. This implies that the dataset contains 392 cases in which the Commission had discretion when choosing the legal instrument. To measure the degree of euroscepticism, we calculated the average position of the Council along the pro-or anti-EU dimension by aggregating the positions of all national governments represented in the Council at the time when the Commission published its proposal. National governments' positions are based on national election manifesto data by Volkens et al. (2018) and constructed, respectively, using the positive (per 108) and negative (per 110) mentions of the EU in the most recent manifesto at the time of proposal publication and weighted by a party's government seat share. Specifically, we subtracted positive from negative statements, which implies that higher values indicate more euroscepticism. We calculated both an unweighted and a weighted version of the euroscepticism variable. The latter variable takes voting weights in the Council into account.
We ran our models both with and without fixed effects for policy areas, which we approximated by the lead committee in the EP dealing with the Commission proposal. Moreover, as the Commission might also pick a given legal instrument for purely functional reasons, we added control variables for the characteristics of the individual policy proposals. In particular, we controlled for three types of complexity associated with any given Commission proposal, following the approach recently advocated by Hurka and Haag (2019): the proposal's structural size (i.e., the number of recitals, paragraphs, subparagraphs, points and indents), its linguistic complexity (i.e. its word entropy (Shannon, 1948)) and its relational complexity (i.e., the average number of cross references per article). As it is often claimed that regulations should be picked in situations of high technical complexity, these variables should be taken into account in order to control for functional considerations.

IV. Empirical Analysis
Before we present our findings, we first provide some evidence to support our general claim that directives typically involve more discretion for the member states than regulations. As our own data did not allow us to perform this analysis, we re-analysed data on delegation patterns in 158 major EU laws assembled by Fabio Franchino (2007). In these 158 laws, Franchino first identified all major provisions and then coded the provisions that delegated power either to national administrations or the Commission. The shares of major provisions including the delegation of power could then be calculated as the delegation ratio of national administrations and the Commission, respectively. Figure 1 plots these delegation ratios and allows us to distinguish between directives and regulations. We found a clear pattern: in general, directives delegate much more power to national administrations than to the Commission and this pattern is the exact opposite for regulations. Thus, what Franchino's data show is that while regulations do indeed sometimes entail some discretion for member states, they do so to a much lower degree than directives. Moreover, regulations primarily delegate powers to the Commission, not to the member states, which underscores the main assumption we make in this article.
In the previous section we hypothesized that policy legacy plays a crucial role in the Commission's choice of legal instruments and that, as a consequence, the original instrument choice will be reflected in subsequent decisions (H1). In order to test this claim, we looked at Commission proposals that either amended or repealed existing EU legislation. If acts were primarily amended or repealed by acts that use the same legal instrument, we considered this evidence in support of our hypothesis. To determine whether a policy proposal amended or repealed another existing piece of legislation, we consulted the EP's Legislative Observatory and coded the required information for each proposal in our dataset. Figures 2a and 2b summarize the required empirical evidence. The thickness of the arrows indicates the relative representation of a given scenario in the data. A range of observations are relevant. As Figure 2a indicates, there are indeed instances when the Commission does not follow the advice of its legal service and amends a legislative act using a different legal instrument. However, these instances are relatively rare. In only 10 to 11 per cent of all acts that amended existing EU law, did the Commission use a Legal Instrument Choice in the European Union different legal instrument and in most of these instances, the respective amendments are rather technical in nature. Most of the time, the amendment simply states that the amended regulation or directive should include a reference to the newly introduced law but does not touch the substance of the amended law. Accordingly, once the Commission has decided to legislate using a given legal instrument, future amendments are usually introduced using the same instrument. Figure 2b provides another test of the argument in H1, depicting the extent to which the Commission sticks with the legal instrument it has chosen in the past when it completely repeals an existing regulation or directive. The most important message here is that while our data contain several instances in which regulations replaced directives, not a single directive has repealed a regulation. Accordingly, once the Commission has decided to enter the path towards EU-wide harmonization by issuing regulations on a certain policy matter there is little chance that the member states will regain discretion in this area in the future. At the same time, the Commission seems more ready to take away discretion it has granted in the past. In general, if the Commission changes its prior instrument decisions, it changes it towards more harmonization and less discretion by member states. However, these instances are also an exception, not the rule. In a large majority of cases the Commission sticks with the legal instrument it has used in the past when it replaces a given piece of legislation. Accordingly, the results broadly corroborate our hypothesis on the legacy of prior decisions, with a few exceptions.
Yet we did not only expect the shadow of prior decisions to determine the Commission's discretion in later policy options, but also to account for the cross-sectoral variance in the choice of legal instruments. Here, we expected the Commission to rely more strongly on directives in areas that have a tradition of being dealt with under the co-decision procedure.  In order to test whether the choice of legal instruments varies systematically across policy areas, we looked at the distribution of directives and regulations in these policy areas. Specifically, we identified the policy area of a proposal by the lead committee in the EP responsible for drafting the EP position. Table 1 displays the results. The table contains several important messages. In general, the Commission always enjoys some discretion when choosing the legal instrument in almost all policy areas. But the extent to which this discretion applies varies. The Commission's hands are bound by the treaties most of the time when it makes proposals in international trade, regional development and constitutional affairs. In all of these policy areas, it almost exclusively issues regulations, because the treaties require it and because it is arguably the most efficient instrument from a functional perspective. In other instances, the Commission theoretically has some leeway in deciding whether to propose a directive or a regulation, but in practice it always opts for regulations. This pattern is most pronounced in the fisheries sector, where the treaties do not prescribe regulations, but the Commission apparently finds them the most suitable legal instruments for addressing these issues.Yet while there are seven policy areas in which the Commission always picks regulations even when it has discretion, there is variance in the remaining ten. The extent to which the Commission chooses a directive varies widely, from below 10% in civil liberties, justice and home affairs, and agriculture and rural development to more than 50% in internal market and consumer protection. Directives are also relatively likely in legal affairs (50%), environment, public health and food safety (44%) and transport and tourism (40%). Thus, there is a set of policy topics in which the Commission is often inclined to grant member states discretion and others in which it rarely or even never does so, even though it could. But why is this the case? Yordanova (2009) proposes a classification of EP committees based on their influence. Here, EP committees are differentiated into those that are more or less influential based on the share of co-decision files adopted since the Treaty of Amsterdam. The committees on budgets, transport and tourism, internal market and consumer protection, legal affairs, employment and social affairs, economic and monetary affairs, environment, public health and food safety, industry, research and energy, civil liberties, justice and home affairs, culture and education are considered powerful, given their traditionally strong involvement in the co-decision process. The committees on budgetary control, agriculture and rural development, foreign affairs, constitutional affairs, development, international trade, fisheries and regional development, by contrast, are deemed less powerful (see also Chiru, 2019). When comparing the share of directives under discretion with the categorization of the different committees, it becomes apparent that all six policy areas with the highest share of directives are those in which the EP has a strong tradition of being involved as a co-legislator. The exact opposite, in turn, applies for the remaining policy areas. The most important exemptions from this general rule are the committees on budgets and civil liberties, justice and home affairs, where we would generally expect more directives, considering their classification as overall powerful committees. While the lack of directives on budgetary issues is not particularly surprising, the findings for the civil liberties, justice and home affairs committee are more puzzling given the very pronounced differences in the share of directives between those proposals for which the Commission had actually no discretion (0.43) and those for which it enjoyed some discretionary power (0.02). Despite these exemptions, however, the share of directives strongly varies with the extent to which the respective policy areas had previously fallen under the co-decision rule and thus whether or not the EP has been involved in the legislative procedure. This confirms our second hypothesis (H2).
We tested the extent to which the anticipation of the ideological orientations in the Council drives the Commission's choice of legal instruments by means of logistic regressions. Our dependent variable captures whether the Commission chose to issue a directive (1) instead of a regulation (0). As the previous discussion has shown, amending legislation rarely changes the underlying legal instrument, but this leaves us with the question of why the Commission chooses a certain legal instrument in the first place. We therefore restricted the following analysis to cases in which the Commission had discretion and the proposal was new, that is, not amending existing legislation. This left us with a total of 210 observations. We tested our competing hypotheses on the impact of euroscepticism (H3a and H3b) by comparing the odds of choosing a directive over a regulation by analysing the effect of mean euroscepticism in the Council. As shown in Model 1 in Table 2, we see a significant reduction in the odds of proposing a directive if euroscepticism increases. Accordingly, the Commission has not reacted to increasing euroscepticism in the Council by a stronger reliance on directives, but by issuing more regulations, supporting H3b over H3a. These effects remain robust even if we control for whether or not the Commission might have focused their legal activities on issue areas for which the use of regulation is generally more likely by including a fixed effects estimator into the analysis (Model 2 in Table 2). This essentially also controls for the mechanism we proposed in H2. The pronounced increase in the pseudo R 2 statistic indicates that a lot of the variation in the choice of legal instruments is explained by these fixed effects for policy areas. To evaluate the substantive effect, Figure 3 plots the predicted probabilities of choosing a directive at different levels of (unweighted) euroscepticism (the predictions are based on Model 2). The model predicts the probability of a directive at 36 per cent when unweighted euroscepticism is at its minimum, whereas the predicted probability is only seven per cent when euroscepticism has its maximal value. It is interesting to note, however, that the effects of euroscepticism disappear if we weigh the national governments' positions with their voting power in the Council (models 3 and 4). The weighted and the unweighted versions of the euroscepticism variable differ markedly from each other, as eurosceptic governments have mainly taken office in small member states during the time period we analyse, with the major exception of Italy in May 2018. This means that unweighted euroscepticism has increased over time, whereas weighted euroscepticism oscillated around a constant mean until the Italian government took office. As the Council of Ministers operates on a strong consensus norm (Heisenberg, 2005), however, the unweighted euroscepticism variable might be more informative of the actual degree of euroscepticism in the Council than the weighted version. In other words, even though small governments like Austria have comparably little voting power, the fact that the Council often aims to incorporate as many views as possible when formulating its positions implies that also small governments can have an impact on Council deliberations, even if the final vote takes place under qualified majority rule. Another possible explanation for these contradictory findings may be that the increasing euroscepticism in the Council has been accompanied by the transition from the Barroso II to the Juncker Commission (since November 2014). In direct response to the decreasing legitimacy of the EU and the rise of eurosceptic populist movements, Juncker announced that the EU should become 'bigger and more ambitious on big things, and smaller and more modest on small things' (Kassim, 2017, p. 15). As found by several scholars, this guiding principle led to a sudden decline of the overall legislative activity by the Commission from the year 2014 onwards (Ząbkowicz, 2018). It has been the introduction of project team leading vice-presidents that specifically allowed the presidency to exert a stronger influence over the Commission's agenda and thus to take account of the  (Bürgin, 2018). Our findings correspond to these insights, suggesting thatwhile the Commission seems to have done overall lessit has done so while providing less leeway to member states when it considers legislative action necessary. It thus appears as if the Juncker Commission has met the Eurosceptic tendencies in the Council and beyond both by streamlining its legislative agenda and by reducing the member states' room to manoeuvre when initiating new legislation. A potential reason for this may be that the Commission is keen to restrict member states' discretion to dilute supranational legislation at the transposition stage in an evermore eurosceptic environment.

Conclusion
If the Treaty does not provide clear legal guidance, the question of whether the Commission highlights the need for strong harmonization through regulations or the need for discretion by member states through directives is highly political, as the recent controversy around the Copyright Directive has demonstrated. Given the political nature of the decision, it is surprising that legal scholars, not political scientists, have mainly focused on the choice of legal instruments. While existing legal scholarship provides valuable descriptive accounts of the legal scope conditions in which the institutions operate, the crucial question of what ultimately drives the legal instrument choice by the Commission has been neglected by scholars of European integration.
Building on novel data from the post-Lisbon period, we show that the choice of the Commission (1) is highly determined by prior policy decisions, (2) varies systematically across policy areas and is (3) sensitive to the ideological orientations within both the Commission and the Council. Interestingly, the Commission did not respond to rising euroscepticism by increasing leeway for member states in the transposition phase. On the contrary, the Commission has shown a tendency to keep member states on a short leash in recent years by issuing an increasing amount of regulations. This resonates well with previous findings which show that the Commission resorts to the adoption of tertiary acts when the danger of legislative gridlock increases (Junge et al., 2015). Apparently, the Commission shifts its legislative efforts both vertically and horizontally when the constellation of institutional actors becomes more adversarial. Both these moves towards more tertiary acts and towards more regulations can be interpreted as indicators that the Commission seeks to strengthen its grip on the adoption and enforcement of EU law when the decision-making process becomes more challenging.
What are the broader implications of these results for policy-making in the EU? First, our findings strongly suggest that in many instances, the Commission is severely constrained in its ability to deviate from the legislative trajectory on which it has started in the past. This implies that whenever the Commission proposes legislation on a new policy issue, the legal instrument it chooses determines not only the leeway granted to member states in the present, but also in the future. In other words, once the Commission has made the case that the approach towards a given policy problem should be harmonized across the EU, future Commissions will find it extremely hard to argue that member states should be granted more discretion on that matter. It is somewhat more likely that the Commission will withdraw the discretion it has granted in the past. But in general, the initial decision to use a certain legal instrument strongly determines future instrument choices.
Second, the Commission's tendency to resort to more regulations in the face of rising euroscepticism raises a crucial normative question: is more Europe, in the sense of 'more harmonization', the appropriate answer to the populist threat or do we need to leave more room for eurosceptic governments to choose their policy instruments without pressure from the supranational level? While we cannot give a definitive answer to this question, the Commission has clearly favoured the first view in recent years. Yet it is unclear how sustainable this path will be. As our data suggest, the Juncker Commission's choice to rely more strongly on regulations will also bind future commissions that may be faced with even more eurosceptic national governments.
Future research should complement the quantitative approach we took in this article by providing qualitative evidence on the precise justifications brought to bear by the Commission when it proposes new legislation. As our introductory example on the Copyright Directive and the General Data Protection Regulation has shown, the Commission is often quite arbitrary when it justifies its choice of the legal instrument. A systematic assessment of the explanatory memoranda that precede the legal text of the Commission proposal and interviews with policy-makers in the Commission would help shed further light on the concrete motivations that guide the Commission when it selects a legal instrument.
On a general level, our study highlights the necessity for EU scholarship to engage systematically with the legal specificities surrounding individual legislative procedures. While institutional and political factors are clearly important for understanding how the EU functions, political scientists should not leave the analysis of the legal aspects exclusively to legal scholars. Especially in times of rising euroscepticism, the choice of a legal instrument is only partially a result of functional considerations. It is also a political question.