‘I did it my way’: customisation and practical compliance with EU policies

ABSTRACT

By tackling shared problems through concerted policies, the European Union (EU) is thought to have a superior output legitimacy. However, EU policies change as they are being ‘customised’ during the implementation process. How do such patterns of ‘differentiated implementation’ affect EU governance in practice? While some studies highlight the danger of ‘watering down’ the objectives of EU law, others emphasise the role of decentralised problem-solving. We analyse how customisation affects states’ practical compliance with EU anti-discrimination, environmental, and justice and home affairs directives in 27 member states (excluding Croatia) between 2007 and 2013. The findings show that customised density (higher number of rules than prescribed by the EU directives) reduces practical compliance. Conversely, customised restrictiveness (stricter requirements than the EU directives) improves practical compliance. In contrast to earlier implementation research, we conclude that literal implementation is not the best form to ensure practical implementation.

KEYWORDS:

• Customisation
• differentiated implementation
• Europeanisation
• gold-plating
• over-compliance
• practical compliance

Introduction

Given relatively low levels of democratic inclusivity in European Union (EU) politics, its capacity to effectively achieve policy outcomes is a key source of legitimacy for citizens (Majone 1999; Scharpf, 1999; Schmidt, 2013; Strebel et al., 2018). It is therefore surprising that the output legitimacy of EU governance receives much less empirical scholarly scrutiny than its input side. Underlying its unique multilevel governance structure is the expectation that EU policies will induce changes in the behaviour of decentralised implementers, which are needed to solve shared policy problems.

Moreover, member states routinely interpret and change EU policies when transposing them into national law, a process that we refer to as customisation (Fink & Ruffing, 2017; Logmani et al., 2017; Thomann, 2019). For instance, Directive 2001/51/EC requires member states to prosecute carriers who transport illegal immigrants into EU territories. However, not all countries explicitly ensure that imposed sanctions do not decrease the protection of asylum seekers; and many governments only foresee a subset of possible sanctions. Beyond legal compliance with EU policies, these more fine-grained, diverse patterns of implementation are widespread but under-researched (Thomann & Zhelyazkova, 2017).

Furthermore, as Treib (2014, p. 29) states, ‘we have as yet comparatively little evidence on the extent to which there is non-compliance beyond transposition and on the factors that are conducive to effective application and enforcement’. Thus, ‘a directive can be perfectly transposed into national legislation, but this does not necessarily lead to practical implementation as well’ (Versluis, 2007, p. 51). Empirical studies support the conjecture that legal and practical compliance are ‘decoupled’ (Falkner et al., 2005; Versluis, 2007; Zhelyazkova et al., 2016, 2018).

The empirical question remains, then, how domestic adaptations of EU rules (i.e., customisation) affect their implementation in practice. Just as differentiated integration enables member states to opt in or out of EU policies, it is important to study ‘differentiations’ in national responses to EU rules because, ultimately, any EU policy needs to be implemented by the relevant national institutions (Jordan, 1997). Therefore, this study asks: how do different dimensions of customisation affect states’ practical compliance with EU law?

Addressing this question contributes to theoretical debates on the interactions between EU legislation, domestic policies and practice. The literature on better regulation treats non-literal interpretations of EU rules as a potential threat as it can create unnecessary red tape, hamper the competitiveness of businesses in the single market, or be a symptom of blame avoidance or blame-shifting (Baldwin et al., 2010; Jans et al., 2009; Radaelli & Meuwese, 2009; Voermans, 2009). Conversely, theories of multilevel governance view customisation as a problem-solving strategy that facilitates adaptations of centrally decided policies to the local context and enhances their acceptance among target groups (Hooghe & Marks, 2003; Matland, 1995).

Our key argument is that in order to test these opposing views, we need to distinguish between customisation as changes in the density or the restrictiveness of EU rules through national legislation. This well-established distinction between different dimensions of changes in regulatory outputs (Knill et al., 2012; Schaffrin et al., 2015) not only informs the measurement of customisation (see Thomann, 2019), but it also matters for understanding its effects. We derive hypotheses that link two distinct dimensions of customisation with practical compliance, while accounting for legal compliance, domestic politics and EU policy characteristics.

Our empirical analysis constitutes the first scholarly attempt at systematic comparative assessment of the relationship between customisation and practical compliance in 27 EU member states (excluding Croatia) and across three policy areas: Justice and Home Affairs (JHA), Social Policy, and Environment. In line with our key argument, we find that whereas customised density increases the likelihood of states’ non-compliance in practice, higher levels of customised restrictiveness are generally conducive to full practical compliance. These findings support theories of multi-level governance underscoring that the practical impact of EU rules is mediated by regulatory differentiations beyond legal compliance.

Legal compliance, customisation, and practical compliance

The question why EU member do (not) comply with EU law has been extensively researched with respect to member states’ transposition of EU Directives into national legislation (Treib, 2014). Correct transposition means that EU directives are incorporated in national legislation in a way that is compatible with the EU objectives (Zhelyazkova et al., 2016) Table 1.

Table 1. Legal compliance, customisation, and practical compliance.

Process	Focus	Actors	Criteria
Legal trans-position (on paper)	Legal & bureaucratic measures for transposing EU law	Legislators / parliaments Interest groups, veto players Administration	Legal compliance Substantive legal conformity of national rules with EU rules Customisation Extent to which national rules differ from EU rule: density (number of rules) restrictiveness (state intervention)
Practical implement-tation (in practice)	National & subnational regulatory practice & styles Organisational & administrative structures	Administrative units Agencies Public servants Policy target groups	Practical compliance Extent to which: activities of administrative actors conform with EU objectives member states implement effective & dissuasive sanctions against non-compliance & efficient court proceedings

However, legal compliance does not capture the full process of translating policy into action (Thomann & Zhelyazkova, 2017; Versluis, 2007). Instead, customisation reflects the more fine-grained patterns of how countries use their discretion to adapt policies to local circumstances during transposition (Thomann, 2019). Customisation differs from legal compliance because it refers to the extent to which member states change the EU rules when transposing them into national law. Even if countries comply with EU rules, they can customise EU Directives to a considerable extent in order to adjust them to domestic contexts (Fink & Ruffing, 2017; Logmani et al., 2017; Thomann, 2019; Thomann & Zhelyazkova, 2017). These adaptations are highly political, as they offer opportunities for member states to influence and shape EU policies in line with their preferences and capacities (Richardson, 2012). Because customisation offers more subtle and often non-punishable ways for member states to ‘have it their own way’, it differs conceptually and empirically from legal compliance (Thomann, 2019).

In essence, customisation refers to vertical policy changes along the implementation chain. When comparing differences between EU and domestic rules, we use well-established approaches to classifying regulatory outputs and temporal and spatial changes therein to measure customisation (Schaffrin et al., 2015; Thomann, 2019). Customisation is defined as the changes that EU rules undergo in their density and restrictiveness during transposition. Added customised density refers to the number of national rules in addition to the EU requirements—the ‘quantity’ of national regulation. For instance, member states can implement several different types of financial sanctions on carriers facilitating illegal immigration foreseen in Directive 2001/51/EC, while the directive requires the implementation of only one type of sanctions. However, adding sanctions does not necessarily increase the restrictiveness of EU legislation, unless member states specify how these should be applied. Additional rules could also decrease the restrictiveness of a policy, if they concern exemptions from the EU requirements, increasig the discretion of target groups.

Whereas customised density does not refer to substantive aspects of regulatory change, customised restrictiveness reflects how domestic rules differ from the EU legislation in content: i.e., the extent to which the state intervenes into and restricts the behaviour of target groups (Knill et al., 2012; Thomann & Zhelyazkova, 2017). For instance, member states can specify the minimum or maximum amount of financial penalties (or both) to be paid by unauthorised carriers of refugees (as prescribed by Directive 2001/51/EC), thus restricting the freedom of maneuver of national bureaucracies to set these limits themselves.

Density and restrictiveness can be added to EU rules (by implementing more rules or formulating stricter requirements than an EU directive). Alternatively, member states can adopt fewer or less restrictive rules. For example, states can choose not to transpose some of the conditions for granting residence permits to third-country nationals or omit certain definitions from the directive (less density). Whereas such omissions can result in legal non-compliance, they do not necessarily exacerbate practical implementation. By partially transposing the EU directives, policy-makers either consciously or unwittingly delegate implementing authorities to national bureaucrats to fill in the policy gaps at the implementation stage. Recent research shows that implementers are more likely to comply with the EU directives in practice when national legislation misses important EU provisions (Zhelyazkova et al., 2018). Some omissions can even broaden the scope of national measures beyond the target groups defined in EU directives.

Contrary to legal compliance and customisation, practical implementation refers to the process in which rules are enforced and applied by the relevant administrative actors and target groups —not through intentions ‘on paper’, but through activities ‘in action’ (Thomann, 2019; Treib, 2014; Versluis, 2007). While practical implementation can result in diverse outcomes (Bondarouk & Mastenbroek, 2018), we seek to explain the practical compliance with EU rules by member states.

The existing empirical evidence about the link between legal and practical compliance is limited and inconclusive (Knill et al., 2012). While it has been suggested that customisation might prove the missing link (Thomann & Zhelyazkova, 2017), the relationship between customisation and practical compliance has hardly been studied empirically. Aggregated customisation levels have a strongly context-dependent effect on practical implementation (Thomann, 2019). We need to theorise both possible positive and negative effects of customisation and control for legal compliance levels as well as country- and sector-specific patterns (Treib, 2014).

Watering down or problem-solving?

The key question of this study is how customisation affects practical compliance with EU rules.

Based on EU’s ‘better regulation’ agenda, reducing ‘unnecessary over-regulation’ among other things aims to ‘increase competitiveness by minimising regulatory burdens and providing efficient regulations’ (Radaelli & Meuwese, 2009, p. 639). This implies transposition should be restricted to what is necessary to comply with the EU’s minimum requirements (Voermans, 2009), where customisation is not desirable.

Two possible mechanisms underlie the negative effect of customisation. First, by extensively changing an EU directive, member states create new rules that essentially differ from the initial EU policy. As a result, national implementers need to consider both the EU directives and national measures (Dörrenbächer, 2017). The increased divergence between the national and EU requirements increases the number of possible implementation alternatives creating ambiguity which policy should be followed in practice. Second, customisation is often associated with red tape ‘resulting in unnecessary burdens and competitive disadvantages for domestic businesses’ (Voermans, 2009, p. 8). Administrative burdens create considerable costs (psychological and others) for actors who implement them (Burden et al., 2012). Target groups and businesses who feel disadvantaged might exert pressure on bureaucrats to implement the rules differently. Moreover, red tape can favour corruption (Dunlop & Radaelli, 2019) and divert resources from effective practical compliance. A lack of perceived meaningfulness or legitimacy of the customised policy might make domestic bureaucrats reluctant to implement these rules (Hartlapp, 2014; May, 2015).

However, case-study evidence suggests only limited negative effects and sometimes positive effects of legal ‘over-implementation’ (Falkner et al., 2005; Jans et al., 2009; Thomann, 2019; Treib 2014; Versluis, 2007; Voermans, 2009). We argue that to better understand the effects of customisation, we should make a distinction between customised density and restrictiveness (Knill et al., 2012; Thomann & Zhelyazkova, 2017). The effects of customisation may vary depending on whether member states add ‘extra requirements’ to an EU rule (Versluis, 2007) (the number of additional national rules relative to the original EU policy), or whether they lay down stricter requirements on target groups (the ‘substance’ of national rules). The effects of these two distinct dimensions of customisation on practical compliance are yet to be empirically tested.

More rules, more problems?

The inclusion of more rules (density) introduces ambiguity in the implementation process. This ambiguity is the result of a control deficit: member states ‘doing things their own way’, by adding special rules or creating exemptions, can lead to a watering down of the original intentions of EU policies (Knill, 2015).

Customised density increases the number of rules that need to be implemented by the relevant national institutions. Consequently, higher number of rules increases the chance that national implementers will fail to implement at least one rule correctly either because of capacity limitations or because the additional requirements deviate from their policy preferences. First, increasing the number of rules enhances the costs of implementation, as national implementers need to take necessary actions regarding the additional requirements or exemptions. As the number of necessary implementation activities grows, the possibility for implementation failure is expected to increase. A higher number of rules also magnifies the number of actors involved in the implementation process, who could have diverging preferences regarding implementation (Baldwin et al., 2010; Sabatier & Mazmanian, 1980). Second, additional rules increase the amount of resources that are needed for implementation and monitoring compliance. National legislators may choose to prioritise monitoring the implementation of some rules and fail to detect and prosecute compliance problems with others. Third, when member states add rules that were not included in the original EU directive, they create uncertainty among national bureaucrats whether they should follow the EU or national legislation (Dörrenbächer, 2017).

Accordingly, added customised density is expected to negatively affect practical compliance, by blurring the original intentions of EU rules and contributing to the growth of regulation (Radaelli & Meuwese, 2009), thus complicating the implementation process.

‘Watering down’ hypothesis (H1):

Higher levels of customised density are negatively associated with practical compliance.

Conversely, increased density could also reflect policy-makers’ incentives to take a more proactive approach to practical compliance, where additional rules serve to ‘stack the deck’ in favour of national preferences. National legislators sometimes add rules to clarify ambiguous EU directives, which can facilitate practical implementation (Knill & Lenschow, 1998). The effect of density on implementation might thus be conditional on the preferences and objectives of governments, who specify the EU rules to ensure proper practical implementation.

We acknowledge the relevance of such indirect effects. However, we focus on the direct relationship between density and practical implementation, while controlling for government motivations and bureaucratic capacities. Whereas additional rules can certainly clarify ambiguous EU directives, regulatory quantity on its own does not elucidate the conditions for practical compliance. For example, when transposing the EU equal treatment directives, many member states added exemptions to the norm of non-discrimination for certain occupations (e.g., police and army). However, without specifying the conditions for difference in treatment, the new rules create ambiguity about the application of these directives and ‘water down’ their core objectives to prevent discrimination in employment conditions.

The benefits of restrictiveness

Customisation can also have positive effects on policy outcomes (Thomann, 2019). Jans et al. (2009, pp. 434–435) even argue that some EU policies are actually intended to be ‘minimal rules’ which member states are allowed to exceed:

the whole system of environmental protection within EC policies is based on the assumption that Member States would make use of Article 176 EC Treaty [which foresees the possibility for Member States to take stricter measures]. (…)

This perspective refers to the content rather than the quantity of regulation, and resonates with the idea of problem-solving in multilevel governance (Hooghe & Marks, 2003; Lodge & Wegrich, 2014; Thomann & Sager, 2017; Trein et al., 2019). Regulatory stringency puts pressure on implementers and target groups to make the EU policies work (Knill et al., 2012; Thomann, 2019). Far from watering down original policy intentions, customised restrictiveness reduces the ambiguity of EU rules by specifying the obligations of national implementers. The additional clarity provided through substantive adaptations can enhance the sense of acceptance among practical policy implementers, and should facilitate practical compliance (Hooghe & Marks, 2003; Knill, 2015):

‘Problem-solving’ hypothesis (H2a)

Higher levels of customised restrictiveness are positively associated with practical compliance.

However, it is also possible that restrictiveness has a negative effect on practical implementation, if more restrictive transposition increases the costs of implementation. For example, by limiting the implementation options, restrictiveness removes the flexibility of national bureaucracies to adjust the EU rules according to their capacities and preferences. Highly specific and precise rules can raise controversies among implementers, who disagree with the EU requirements. Furthermore, some implementers may face difficulties complying with very precise high standards that do not allow for divergent policy outcomes. For example, not all member states are able to curb CO2 emissions to the same degree and flexible transposition allows national implementers to search for alternative solutions to a policy problem.

Therefore, we consider a contrasting hypothesis on the effect of restrictiveness on practical compliance.

‘Limited freedom’ hypothesis (H2b)

Higher levels of customised restrictiveness are negatively associated with practical compliance.

Data and methods

We systematically compare the implementation of EU policies in the areas of JHA (10 Directives), Social Policy (4 Directives), and Environment (3 Directives). Including diverse policy areas enables us to identify sector-specific patterns and increases the external validity of our findings (Gerring, 2008). All three policy areas allow member states to go further than what the EU minimally requires. At the same time, they also reflect different constellations of domestic actors and preferences. For example, environmental policies have individual actors or businesses as their main target group, whereas JHA and social policies primarily require national immigration authorities to take action. To ensure that the results are not driven by sector-specific dynamics, we also tested our hypotheses separately for each area.

Table A1 in the appendix provides information about the EU directives, and the data sources for customisation and practical compliance. We only focus on EU directives because other types of EU legislation (e.g., regulations and decisions) are not subject to national transposition and therefore, do not undergo customisation. Furthermore, we rely on conformity-checking reports prepared by external experts contracted by the EU Commission (Zhelyazkova et al., 2016). Based on these reports, we obtained data on both customisation and practical compliance for 277 observations (directive*country dyads). Furthermore, we remain cautious about generalising our findings, given the small number of Social Policy and Environment directives. Nevertheless, the uniqueness of the data enables a first quantitative analysis of the relationship between distinct dimensions of customisation and practical compliance.

Measurement of practical compliance and method of analysis

To obtain comparable information about member states’ implementation activities we employ a novel data-set on practical compliance whose content and data collection procedure have been extensively described elsewhere (Zhelyazkova et al., 2016). Information about practical compliance was extracted from external evaluation reports prepared by consultancies that aim to provide in-depth analysis of member states’ implementation. Conformity studies cover the implementation performance in both Central and Eastern European (CEE) and EU-15 member states in the period between 2007 and 2013. Thus, the analysis excludes Croatia, which joined the EU in 2013.

The evaluations of practical compliance are specific to the requirements of a particular provision in an EU directive. Whereas Zhelyazkova et al. (2016) measure practical compliance as the share of correctly applied provisions relative to all evaluated provisions in a directive, this study focuses on a smaller number of EU directives, resulting in limited variation and a skewed distribution in the share of practical compliance. Therefore, we distinguish between different categories of practical compliance: low, limited and full compliance. Full compliance reflects situations where national experts did not encounter any practical issues related to the implementation of an EU directive. Limited compliance means that national implementation resulted in one major practical problem, while low practical compliance indicates multiple major implementation problems. The measurement closely resonates with qualitative ideas about practical compliance (Falkner et al., 2005). Furthermore, we also acknowledge that information about practical compliance could be affected by characteristics of the assessment reports. Therefore, the analysis accounts for the number of sources providing information about practical implementation and the length of the reports. Despite these caveats, previous research has shown that external assessment reports are highly comparable, while offering a novel tool to analyse member states’ compliance in a quantitative framework (Zhelyazkova et al., 2016, 2018).

We employ multinomial logit as the method of analysis. Alternative methods, such as ordered logit and generalised ordered logit rely on more restrictive assumptions about the effects of the independent variables across different categories of the dependent variable. As a robustness check, we replicated the analysis using ordered logit and generalised ordered logit models (see online appendix).

Measurement of customisation

Our analysis of customisation is based on conformity reports that contain so-called ‘tables of correspondence’ (TOCs) including the directive provisions, the domestic transposition measure, and an evaluation of a state’s level of conformity. The tables also include detailed information about the nature of transposition outputs and whether they exceed the requirements of a directive (Thomann & Zhelyazkova, 2017).

Customised density (H1) refers to the number of rules (policies and/or instruments) that are added to or removed from an EU provision in the transposition measures (e.g., scope restrictions, requirements, conditions, exemptions). A domestic rule customises an EU provision with more density, if it adds at least one new element in the transposition process. Alternatively, a member state may also remove density from the EU directive. For example, Finland and the Netherlands did not transpose optional provisions from the Directive 2003/86/EC on family reunification that would allow immigration authorities to reject applications for entry and residence on grounds of public health. We counted the number of times that an expert mentioned an addition to (coded as 1) or an omission (coded as −1) from a specific EU provision. Density levels range between −2 and 4 at the provision level. At the directive level, we took the average density score of all evaluated directive provisions which is a continuous variable ranging between −2 and 2.

Customised restrictiveness (H2a & H2b) is based on expert assessments that identified transposition outcomes as either more (coded as 1) or less restrictive (−1; hence, more discretionary to the relevant addressees) than the respective EU provision. For example, external experts indicated that some member states implemented more flexible requirements for processing asylum applications and granted more rights to refugees (e.g., healthcare exceeding emergency cases). Other member states further restricted refugees’ rights relative to the JHA Directives by requesting additional documents for asylum applications. Customised restrictiveness is coded as 0, if a provision was neither assessed as more restrictive nor as less restrictive than the EU provision. The most obvious manifestation of this is literal (equivalent) transposition. Similar to customised density, we operationalise restrictiveness at the directive level as the average score of restrictiveness across all evaluated provisions in a directive (it ranges between −2 and 2).

Other independent variables

Our analysis includes policy-area binary variables to account for differences in practical compliance across EU sectors. Furthermore, we control for directive-level characteristics related to the discretion that EU directives grant national administrations and policy complexity. A directive’s discretion is measured as the share of provisions that grant implementation freedom to member states (i.e., ‘may’ clauses) relative to all directive provisions (Thomson & Torenvlied, 2011). We measure policy complexity as the number of recitals preceding the text of an EU directive (Treib, 2014). We also control for a member state’s level of legal compliance with an EU directive (Zhelyazkova et al., 2016), measured as the share of correctly transposed provisions relative to all relevant provisions in a directive. ‘Relevant provisions’ refer to all articles or sub-articles that were assessed as separate issues in the reports and were evaluated as either correctly implemented or not.

We acknowledge that dimensions of customisation and practical compliance can be influenced by similar domestic factors. For example, states that lack administrative capacity to adopt more restrictive measures during the transposition process are also less likely to comply with the EU rules in practice (Knill & Lenschow, 1998; Treib, 2014). To measure differences in member states’ capacity limitations (Börzel et al., 2010), we employ the ‘Government Effectiveness’ indicator from the Worldwide Governance Indicators Database. ‘Government Effectiveness’ reflects the quality of public and civil services at the year of a member state’s directive transposition. In a similar vein, governments that have incentives to deviate from the EU directives are likely to customise national legislation so that policy outcomes favour their own political objectives (Thomann 2019). Incentives to deviate, however, could also lead to non-compliance (Mastenbroek & Kaeding, 2006; Thomson et al., 2007). Thus, we control for government preferences towards the EU. We first identified the relevant governments and their party affiliations based on data from the reports. Information about governments’ positions towards the EU was obtained from the Chapel Hill Expert Surveys (Bakker et al., 2015).

Another factor that could influence both customisation and practical compliance is the number of veto players involved in the transposition process (Zhelyazkova et al., 2016). A higher number of national policy-makers increases the likelihood that at least some legislators will favour adding extra elements to the EU rules or will oppose more restrictive transposition (Thomann, 2019). Veto players also constrain the ability of national actors to effectively comply with the EU requirements (Börzel et al., 2010; Haverland, 2000). We employ two measures for veto players. First, we count the number of ministers involved in the adoption of the transposition measures based on the assessment reports (Zhelyazkova et al., 2016). Second, we control for the extent to which governments consult with civil society organisations (CSOs) in the adoption of national legislation (CSO consultation). The measure is taken from the V-Dem project (Coppedge et al., 2017) and records the extent to which governments routinely give voice to CSOs in the policy-making process.

Finally, we control for characteristics of the assessment reports that could potentially bias the estimates on practical compliance. For example, longer reports are likely to provide more detailed information. Similarly, the use of multiple sources increases the likelihood that at least one of the assessments will describe compliance problems. Thus, we control for the length of the reports (number of pages allotted to a particular country) and the number of sources regarding practical compliance. Table 2A in the online appendix describes the variables employed in the analysis.

The effects of customisation on practical compliance

We first examine whether there are any patterns in customised density and restrictiveness across member states. Figure 1 depicts the average levels of customised density and restrictiveness. Based on our data-set, there are no significant differences in the extent to which member states customise EU directives regarding the three policy areas. The only exception is Latvia, which shows negative average density. However, most member states on average add extra rules to the EU directives during transposition (Figure 1; left-hand side). Conversely, member states’ tendencies to adopt more stringent legislation are more ambiguous. Whereas Belgium, Bulgaria, Germany, Spain and Italy transpose EU legislation more restrictively, other countries show more diversity in customising the EU directives (Figure 1; right-hand side). Thus, although member states frequently modify the EU rules during transposition, they do so in different ways. Furthermore, we observe more variation in density and restrictiveness within member states than across countries. This finding suggests that customisation practices are issue-specific and are driven by different sectorial logics within the EU member states.

Figure 1. Average levels of customised density and restrictiveness across member states.

Table 2 presents the analysis of the effects of customised restrictiveness, and customised density on practical compliance. Models 1 and 3 estimate the likelihood of limited (i.e., only one major gap in practical compliance) relative to full practical compliance (reference category). Models 2 and 4 compare low practical compliance (multiple major practical problems) to full conformity with the EU directives.

Table 2. The effects of customised restrictiveness and density on practical compliance.

	M1 Limited compliance	M2: Low compliance	M3: Limited compliance	M4: Low compliance
	Coef (S.E.)	Coef (S.E.)	Coef (S.E.)	Coef (S.E.)
Density (added rules)	−0.02 (0.26)	0.64* (0.28)	0.31 (0.32)	0.98* (0.41)
Restrictiveness	−0.11 (0.21)	−0.66* (0.32)	−0.42 (0.33)	−1.02* (0.40)
Justice & Home Affairs (=1)			0.57 (1.35)	−2.45 (2.02)
Social Policy (=1)			−0.62 (0.67)	−2.62*** (0.79)
Legal compliance			−0.20 (0.96)	−4.25** (1.31)
Bureaucratic efficiency			−0.42 (0.34)	−0.25 (0.40)
Government EU support			−0.27 (0.25)	−0.19 (0.29)
Directive discretion			−2.68** (0.92)	0.21 (1.12)
Directive complexity (N recitals)			−0.02 (0.03)	−0.07^+ (0.04)
CSO consultation			0.58** (0.19)	0.25 (0.34)
Number of ministers			0.33** (0.12)	0.33* (0.15)
Length of the report			0.02 (0.06)	0.15* (0.07)
Number of sources			0.57 (0.75)	0.07 (1.00)
Constant	−0.78*** (0.17)	−1.15*** (0.26)	−0.54 (3.49)	4.36 (4.17)
Observations	272		236
Pseudo R^2	0.024		0.151
LR chi2	−270.957		−208.917

Notes: Reference category: Full compliance; Policy area reference category: Environment; Standard errors in parentheses (clustered in directives); ⁺ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.

In line with H1 and H2a, the findings suggest that separate aspects of customisation – density and restrictiveness –have distinct effects on practical compliance. More precisely, higher levels of customised density increase the chance that EU member states will experience multiple problems at the practical implementation stage. Conversely, the likelihood of low practical compliance decreases, when national policy-makers transpose the EU rules more stringently. Restrictiveness and density do not significantly affect the likelihood of limited practical compliance.

The estimates support the ‘watering down’ hypothesis that higher levels of customised density significantly increase the likelihood that member states will encounter multiple issues during the implementation process. To better illustrate this finding, Figure 2 (left) shows the changes in the probability of full and low practical compliance (more than one implementation gap) at different levels of customised density. The figure shows that when member states omit (on average) two rules from an EU directive during transposition, the probability of full compliance is 0.74, while the probability of low compliance is only 0.06. This situation pertains to the Latvian transposition of directive 2001/81/EC on emission ceilings for certain atmospheric pollutants. By excluding requirements that pollutants should have ‘adverse effects on the environment’ beyond a critical point, the Latvian government increased the scope of the directive. This omission facilitated implementation, as national authorities could forbid all air pollutants in sensitive areas, instead of having to continuously monitor whether emissions reach critical values. Conversely, the probability of full practical compliance decreases to 0.29, when member states include two additional rules in the transposition of an EU directive, while the chance of encountering multiple implementation gaps increases to 0.49. In particular, when transposing directive 2001/51/EC Poland and Spain added three kinds of financial penalties for carriers of illegal immigrants. The higher number of rules, however, creates confusion about the circumstances under which either of the three penalties would apply, making it more difficult to enforce the rules in practice. This finding is robust across different policy sectors and model specifications (see online appendix). It suggests that concerns about blurring the objectives of EU policies and unnecessary regulation are justified, when examining the amount (density) of national rules.

Figure 2. The effect of customisation (density and restrictiveness) on the probability of full and low practical compliance.

Conversely, the estimates in Table 2 show a significant negative effect of customised restrictiveness on the likelihood of practical non-compliance (low versus full compliance). In line with the problem-solving hypothesis (H2a), customised restrictiveness is conducive to full practical compliance. Instead, we do not find evidence that restrictiveness curbs practical implementation either because it limits bureaucratic discretion or because more concrete rules increase the costs of compliance (H2b). Figure 2 (right) visualises this result. In situations where transposition measures are laxer than the EU requirements, the probabilities of full and low compliance (more than one implementation gap) are 0.23 and 0.53 respectively. These probabilities reflect the less restrictive transposition of Directives 2000/43/EC (on racial equality) and 2000/78/EC (on employment equality) by several member states. In particular, national policy-makers allowed difference in treatment based on the nature of the occupation (e.g., army, church). While such exemptions are in line with the directive’s provisions, it is unclear which types of occupational characteristics justify unequal treatment in practice. Conversely, the probability of full compliance increases to 0.74, when national transposition is more concrete than the EU directives, while the probability of encountering multiple implementation issues decreases to 0.07. For example, the French government adopted more stringent rules than Directive 2006/54/EC on gender equality, by specifying the obligations of employers and employees to negotiate on the pay gap. This result supports our conjecture that EU directives create ambiguities for national implementers that can be resolved by specifying the requirementsdomestic actors need to fulfil.

The analysis supports the findings from previous research on EU compliance. Member states that legally comply with EU directives are less likely to experience multiple practical problems. Furthermore, member states comply better with Social Policy than with Environment directives (Zhelyazkova et al., 2016). Conversely, the number of ministers involved in the transposition process increases the likelihood of non-compliance at the implementation stage. This result is in line with existing research on the negative impact of veto players on member states’ compliance (Börzel et al., 2010; Haverland, 2000; Steunenberg, 2006).

The analysis also shows that directive-level discretion and CSO consultation increase the chance of limited compliance, but they do not significantly affect the likelihood of multiple implementation gaps. The results underscore the ambivalent effect of discretion on implementation. On the one hand, higher discretion facilitates compliance by increasing the range of possible compliant national policies. On the other hand, discretion creates uncertainty about the most appropriate policy alternative from the pool of available options. Similarly, we expected that CSOs act as veto players during implementation. However, CSO consultations could also help governments tailor national transposition measures to the demands of relevant target groups, increasing the chance that the latter will comply with the transposed measures.

Robustness analysis

In the online appendix, we present descriptive analysis (Figure A1) and findings for each separate policy area (Figures A2 & A3) and for different values of legal compliance and government EU support (Table A4). These robustness checks show that customised density and restrictiveness have direct impact on practical compliance that is unimpeded by the political considerations of national legislators. Moreover, the results remain robust for different measures of practical non-compliance (e.g., number of practical problems) (Table A5) and when assuming ordered outcomes for practical non-compliance (Table A6).

Differentiated EU policy implementation: ways ahead

Perspectives on differentiated EU integration typically rest on the assumption that member states achieve compliance because the EU rules provide for sufficient flexibility. However, the phenomena of compliance and customisation have never been studied in conjunction. Past research suggests that levels of legal compliance insufficiently explain practical implementation in the EU (Versluis, 2007; Zhelyazkova et al., 2016). We, therefore, analysed how regulatory changes at the transposition stage affect the implementation of EU directives in practice. In particular, our analysis shows that customisation matters for practical compliance. It is expected that fine-grained patterns of vertical policy change are an essential element of problem-solving in the EU policy-making process (Lodge & Wegrich, 2014; Richardson, 2012; Thomann & Sager, 2017; Trein et al., 2019).

Our findings suggest that the impact of customisation on practical compliance depends on the nature of change: whether member states customise the density or the restrictiveness of EU rules. More precisely, theresults show that different dimensions of customisation have distinct and contrasting effects on practical compliance. For example, adding rules to the EU policies (customised density) tends to adversely affect practical compliance. Thus, higher number of domestic rules indeed complicate practical implementation (Baldwin et al., 2010; Dunlop & Radaelli, 2019; Versluis, 2007). Conversely, we find that customised restrictiveness helps improve practical compliance with EU policies. EU member states that change the content of rules during transposition effectively solve shared policy problems (Thomann & Sager2017; Thomann, 2019). These empirical insights are important because the (intuitively appealing) idea that policy implementation in the EU should generally stay as close as possible to the EU template currently informs the EU better regulation agenda (Radaelli & Meuwese, 2009). Yet, as our results show, this idea is empirically unfounded. Although legal compliance tends to reduce practical compliance problems, the results of this study suggest that EU rules do not directly translate into implementation practices. Thus, in contrast to earlier research, we conclude that literal implementation is not always the best form to ensure practical implementation.

Moreover, the analysis identified a non-linear relationship between customisation and practical compliance. More precisely, customisation alleviates (restrictiveness) or enhances (density) only high levels of practical non-compliance (when member states encounter multiple problems during the implementation process). However, customised restrictiveness does not completely eliminate practical non-compliance. Whereas restrictiveness reduces the likelihood that member states experience multiple compliance problems, it does not eradicate the occurrence of an implementation problem (limited compliance). Similarly, added regulatory density is only associated with multiple practical problems. These findings stress the intricacies of practical implementation, where dimensions of customisation are only one out of many explanations for (non-)compliance. In reality, practical problems occur due to various factors and future research should shed more light on the multitude of explanations that lead to the occurrence of practical non-compliance. We should also bear in mind that practical compliance is not yet a policy outcome. The relationship between practical compliance and behavioural changes of societal target groups is a largely unexplored territory.

Even more than legal compliance, practical compliance with EU policies is a prerequisite for ensuring policy effectiveness and the output legitimacy of EU multilevel governance (Scharpf, 1999; Schmidt, 2013). Our results show that customisation can both impede and contribute to effective problem-solving in the EU. Differentiated implementation’ (Fink & Ruffing, 2017) can lead to either watering down EU policy objectives, or to making European integration a success in practice. Therefore, the outputs of EU governance deserve more empirical scrutiny (Majone 1999; Strebel et al., 2018). Future research should address not only how customisation affects practical compliance, but also the link between differentiated integration, differentiation at the level of transposition, and differentiation at the level of practical implementation—beyond legal compliance.

Acknowledgements

We are grateful to Eva Ruffing, the participants of the panel ‘United in Diversity? Implementation and Performance in Multi-Level Governance’ at the ECPR General Conference, Hamburg, 25.8.2018, the participants of the Public Policy Seminar of the London School of Economics, University College London, and King’s College London, 10.12.2018, and the two anonymous JEPP referees for their valuable comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Asya Zhelyazkova

Asya Zhelyazkova is an Assistant Professor in Comparative and European Public Policy at the Department of Public Administration and Sociology, Erasmus University Rotterdam.

Eva Thomann

Eva Thomann is a Senior Lecturer in Politics at the Department of Politics of University of Exeter.