Content uploaded by Asya Zhelyazkova
All content in this area was uploaded by Asya Zhelyazkova on Oct 01, 2015
Content may be subject to copyright.
The successful transposition of different provisions by member states: An
application to the Framework Equality Directive
Asya Zhelyazkova, Department of Sociology / ICS: Utrecht University
René Torenvlied, Department of Sociology / ICS: Utrecht University and University of
The present study aims to explain variation between member states in compliance with provisions
of an EU law. Predictions are derived about the effects of technical fit, discretion, Commission
warnings, and conflict in the Council on the probability of member-state transposition success
with separate EU policy requirements. Hypotheses are tested on the level of compliance of 15
member states with 27 major provisions laid down in the Framework Equality Directive
(2000/78/EC). Intensive analysis of documents and reports from key informants provided
information on member-state transposition progress at the end of 2004 and 2006. Results show
that the domestic adaptation costs to a provision play an important role in member state
transposition success: high levels of fit and provisions granting discretion improve member-state
transposition success. Formal warnings by the Commission lead to a better compliance record,
while conflict in the Council does not affect the successful transposition of provisions.
Key words: non-compliance, provision-level differences, transposition, goodness-of-fit,
Policy implementation is generally understood as the transmission of the outcomes of
collective decision making into implementers’ actions. Ideally, a measure of compliance
with European Union (EU) policies should include the appropriate actions by national
policy-makers, and even the behaviour by street-level bureaucrats delivering services to
citizens (O’Toole, 2000).
The present study aims to describe and explain variation in the transposition success
of member states regarding specific provisions of an EU law for different periods of time.
Transposition success is defined as the extent to which national implementation measures
are in congruence with the requirements of EU policies. Currently, studies of EU
compliance use multiple indirect indicators for member state performance in the
implementation of EU laws (Hartlapp and Falkner, 2009). Many studies use infringement
procedures instigated by the Commission as an indicator for non-compliance (Mbaye,
2001; Perkins and Neumeyer, 2007). However, this measure is subject to the strategic
behaviour of the Commission and may be biased (Börzel, 2001). Other studies use delays
in the transposition of EU directives as an indicator for non-compliance (Mastenbroek,
2003; Kaeding, 2008; König and Luetgert, 2009). However, transposition rates refer to
member states’ notification reports to the Commission regarding national laws (Hartlapp
and Falkner, 2009). There is no guarantee that the national laws reported indeed
transpose the directives adequately.
One important exception is the study by Falkner et al. (2005) on the implementation
of EU social policy by member states. The authors apply the criterion of correctly
transposed legislation to indicate the moment in time at which there exists a complete
match between the policy intent as expressed in a directive and the actions of a member
state. This approach has the advantage that the exact timing of transposition success can
be verified for each member state. However, the timing of correct transposition does not
inform us about developments in compliance before member states fully complied with a
directive. Moreover, their measure defines compliance with the entire directive. Yet, the
prescribed activities of member states are specified in the different provisions of a
directive, and member states might comply with certain provisions, while failing to
comply with others. Thus, a measure of compliance with EU provisions is more
appropriate (Thomson, 2009b).
In the present study, we introduce the concept of transposition success as a new
indicator for compliance. We build on the work of Falkner et al. (2005) about the
correctness transposition measures by: (1) moving the analysis to specific provisions of a
directive and (2) studying member states policy developments at different points in time.
We introduce a theoretical framework based on expected costs of compliance and non-
compliance with provisions of EU directives. Costs of compliance and non-compliance
do not only vary across member states and directives, but also across member states and
specific provisions within a directive. For example, member states might have established
laws and practices that are compatible with certain aspects of an EU law, but not with
others. In addition, the provisions of an EU law could vary with respect to the discretion
they grant to national implementers.
To test hypotheses, the present study combines qualitative data collection with a
quantitative research design. More precisely, we rely on elaborate content analysis of
numerous implementation reports about the transposition of the Framework Equality
Directive (2000/78/EC) by 15 member states. Case studies on the implementation of this
specific directive focus mostly on only one or few member states and their compliance
with the directive as a whole (Dimitrova and Rhinard, 2005; Falkner and Treib, 2008;
Casado Asensio, 2008). By contrast, we apply statistical analyses to a single case by
moving to a lower level of analysis: the provision level. Such a research design allows for
an in-depth comparison between different aspects of a directive and member states, while
holding constant characteristics at the directive-level – thus locating our research between
large-n and case study approaches.
Member states’ compliance record with EU requirements has attracted increasing
academic attention over the last decades (Mastenbroek, 2005), giving rise to numerous
hypotheses about possible explanatory factors. One main hypothesis, however, has
received broad scholarly acceptance: the hypothesis that non-compliance is only a
temporal phenomenon. Sooner or later member states are supposed to transpose an EU
law correctly due to supranational pressures by the Commission and the ECJ and
domestic influence by national (sub-)actors with vested interests (Héritier et al., 2001).
However, national policy-makers often experience problems when implementing
specific EU requirements, which could be left unnoticed by supra(national) organizations.
Transposition problems can become visible at the domestic level only after national court
cases result into decisions that are inconsistent with the EU requirements. Furthermore,
extensive supranational scrutiny of all national transposition measures is a time-
consuming and costly process, which depends on the resources and policy considerations
of the Commission (Hartlapp and Falkner, 2009). We view the successful implementation
of an EU rule as a function of the costs for meeting an EU rule and the expected
consequences of disobeying it. Thus, we use costs as a heuristic to introduce factors that
affect the transposition of EU provisions across member states.
One of the most widespread hypotheses on EU implementation is ‘the goodness-of fit
hypothesis’: member states’ transposition success depends on the level of compatibility
between the requirements in European directives and existing national policies and
institutional arrangements. The logic is that the goodness-of-fit affects the costs of
adaptation for member states to comply with the European requirements (Héritier, 1996).
Public authorities might not be inclined or able to bear these costs (Börzel, 2003; Falkner
et al., 2005).
The goodness-of –fit hypothesis has been criticized by scholars on the grounds that it
is under-theorized, lacks proper specification, and has limited explanatory power in
empirical analysis (Knill and Lenschow, 1998; Mastenbroek, 2005). Some scholars have
modified their theoretical arguments and claim that misfit is necessary, but not sufficient
to explain non-compliance. Thus, scholars include auxiliary factors, such as veto players
(Haverland, 2000) or the preferences of domestic decision-makers (Steunenberg, 2006).
Others have dismissed the misfit hypothesis as irrelevant (Mastenbroek and Kaeding,
2006) because of lack of empirical support.
We follow the critique that the goodness-of-fit hypothesis remains underspecified
because scholars do not clearly spell out the nature of the costs that emerge from the
incompatibility between European directives and domestic regulations. A European
directive incurs different types of costs on member states and (sub)national actors
depending on the objectives of the directive and the characteristics of member states. As
a result, costs are highly contextual, and depend on the nature of the EU directive. For
example, the implementation of EU anti-discrimination directives incurs normative costs
on domestic actors who do not identify with the norms as stipulated in a directive (see
Dimitrova and Rhinard, 2005). A different type of costs is associated with the legal-
administrative costs for public authorities to design laws that are both compliant with the
EU directive and do not disrupt related domestic structures. Such costs could be referred
to as ‘technical misfit’.
The present study focuses on such legal-administrative costs
created by the incompatibility between national legal texts and legal practices, and each
specific provision of the Framework Equality Directive.
By shifting the unit of analysis from the directive to the provision level, we further
specify the relation between ‘goodness-of-fit’ and compliance in two respects. First,
existing case studies on the relation between misfit and compliance often focus on only
few important provisions of a directive and draw conclusions about the level of misfit for
the whole directive (Haverland, 2000; Bailey, 2002). This approach is problematic as it
leads to a selection bias, since these important provisions reflect the most politicized
issues in a directive. However, directives generally consist of numerous technical
provisions that address different legal aspects and methods to achieve the goals of the
directive. These provisions are not likely to be highly politicized, but could still lead to
substantial differences in judicial interpretation.
Second, provisions contain more specific policy requirements than directives as a
whole. A focus on the provision level helps specify in more detail the nature of the costs
incurred on implementing actors. For provisions that refer to concepts and definitions of
discrimination (i.e. direct and indirect discrimination) the costs are defined in terms of
their legal compatibility with the existing definitions in domestic systems and whether
these definitions are interpreted accordingly in national courts. Even the smallest
adjustment of a definition in a national law could lead to major changes in numerous
national documents. For provisions that refer to sanctions against discrimination and
procedures fighting existing discrimination practices, costs relate to the necessary
changes in the practical implementation of directives by (sub)national actors (Bailey,
2002). These provisions might require changes in the legal procedures, practices, and
routines of (sub)national actors if sanctions and procedures are not sufficiently “effective,
proportionate and dissuasive”. In both cases, expert evaluation reports help identify the
legal consequences for member state governments to introduce laws that comply with
specific provisions. We formulate the following hypothesis:
H1: Member states are more likely to comply with provisions that are more compatible
with existing domestic legal documents, institutions and practices.
Member states’ adaptation costs are also influenced by the level of discretion laid
down in an EU provision. Executive discretion refers to the granting of discretionary
powers to national agencies charged with implementing a decision. Higher levels of
discretion increase the discretionary boundaries around the decision outcomes contained
in the directives (Epstein and O’Halloran, 1999). Thus, when higher levels of discretion
are granted to member states, wider ranges of policy performances are compatible with
the decision outcomes contained in a policy (Steunenberg, 2006; Thomson et al., 2007).
In other words, we expect that the level of discretion decreases the costs of adaptation to
a specific EU provision and consequently increases the likelihood of compliance.
H2a: Member states are more likely to comply with provisions that grant discretion to
Furthermore, one could expect that discretion affects the relationship between
technical fit and transposition success. More precisely, provisions that allow for a number
of implementation outcomes are expected to facilitate transposition especially because
national implementers can introduce few or no changes and, thus, preserve existing
legislation and practices. Thus, member states that lack established policies on a
provision could choose to implement an alternative that is closest to the specific national
context. As a result, discretion downplays the differences between lower and higher
levels of technical fit on member states’ transposition success. By contrast, the effect of
fit is expected to be stronger for provisions that constrain the choice of national
H2b: The difference in compliance between lower and higher levels of technical fit is
larger for provisions granting limited discretion than provisions granting extensive
discretion to member states.
Scholars generally predict a negative relation between discretion and delays in
transposition by member states (Thomson et al., 2007; Kaeding, 2008; Steunenberg and
Toshkov, 2009). The argument is that higher levels of discretion oblige member states to
choose between different policy alternatives to be implemented, which can give rise to
domestic conflict and subsequently delays in the transposition process. However, at least
in the long run, delays in transposition are not necessarily related to the correctness of
transposition measure, which is the main focus of the present study.
The Commission and infringement procedures
Theories on enforcement view compliance as a function of the expected costs from
the likelihood of detection and credible threat of sanctions. As a result, rigorous
monitoring and formal sanctions are expected to increase the costs of non-compliance. In
the European context, increased rigor in the Commission monitoring is reflected by
infringement procedures (Tallberg, 2003). The Commission has extensive powers to issue
warnings to member state governments with a questionable implementation record, and
to pursue formal infringement procedures before the European Court of Justice (ECJ) in
cases of persistent non-compliance. Given that the Commission has the power to propose
punitive fines against member states that violate EU law, court litigations increase the
costs of non-compliance for member states as the threat of sanctions becomes imminent.
Studies also report a negative relation between the instigation of infringement
proceedings and delays in transposition especially after the initial stage of sending ‘letters
of formal notice’ (Zhelyazkova and Torenvlied, 2009).
H3: Member states that received a warning from the Commission are more likely to
comply with the provisions of a directive.
Conflict in the Council
We expect that conflict in the Council affects the transposition success of national
governments. Implementation theory stresses that debated issues can result in ambiguous
and incoherent decision outcomes which constrain the efficiency of implementers (Hill
and Hupe, 1992, McCubbins et al., 1989; Torenvlied, 2000). In addition, the more
conflict exists during decision-making, the more likely it is that the interests of some
member states will not be incorporated in the decision outcome (König and Luetgert,
However, in the European context conflict in the Council increases the vigilance of
the Commission with respect to problem-prone policy issues. As a result, the
Commission could allocate resources to manage and enforce compliance by member
states through informal bargaining, dispute-settlement strategies and credible threat of
sanctions (Tallberg, 2003).
In addition, member states can commit to the transposition of controversial issues
when unequal practices can lead to negative externalities to all member states (Majone,
2001; Franchino, 2007). Based on these considerations, Zhelyazkova and Torenvlied
(2009) predict and find a positive effect of conflict in the Council on member states’
compliance with EU directives.
We, therefore, expect that conflict in the Council on a particular provision has a
positive effect on member states’ transposition success.
H4: Member states are more likely to comply with provisions that were debated during
the Council meetings.
The dataset used in the present study was compiled on the basis of implementation
reports that evaluate the transposition success of member states with each of the
provisions of the Framework Equality Directive (2000/78/EC). This directive was
adopted with the purpose to put in place a general framework to ensure equal treatment of
individuals in the EU, regardless of their religion or belief, disability, age or sexual
orientation, as regards to access to employment or occupation and membership of certain
organizations. The Framework Equality Directive has a number of characteristics that
make it highly suitable for the present study. First, it touches upon a controversial topic
and, hence, had at least some potential to cause problems during the implementation
process. Second, it covers different policy issues related to both policy content and
enforcement mechanisms, which allows testing provision-level hypotheses.
In order to obtain information about the level of compliance of member states with
each separate provision of the directive, we first used information from three extensive
reports published by the European Trade Union Institute (ETUI) at the end of 2004. The
data in the ETUI reports are based on responses from trade unions in the member states to
a questionnaire regarding the transposition of the directive in their national contexts. In
addition, we supplemented the information with expert interviews with the authors of the
reports to ensure a proper understanding of the implementation process. The ETUI
reports and expert interviews provide an important first step to assess the transposition
performance of member states with the provisions of the Framework Equality Directive
by the end of 2004.
The second major source of information is the various publications from the
European Network of Legal Experts (ENLE). These publications consist of country-
specific and more general comparative evaluation reports on member states’ progress
with the implementation of anti-discrimination directives. The reports cover the changes
made to national law and the development of enforcement mechanisms in accordance
with the directive requirements until 7 January 2007. Based on the information provided
by ENLE, we were able to cross-validate the data collected from the ETUI reports until
the end of 2004 and supplement the existing dataset with information on member states
compliance until the end of 2006.
Dependent variable: measurement
We identified the compliance problems that member states experienced during the
implementation of the provisions of the Framework Equality Directive. In this study, a
provision is defined as an entire article or a numbered sub-article of a directive with
which member states should comply through appropriate national legislation. This
encompasses almost all provisions in the directive with the exception of the standard
provisions that relate to the deadline of transposition and the Commission report. We also
excluded one specific provision pertaining only to Northern Ireland. In total, we obtained
data on the transposition performance of 15 member states regarding 27 provisions of the
Framework Equality Directive.
Member states’ transposition performance was coded to indicate whether a member
state fully complied with each of the provisions (coded as 1). If the reports explicitly
mentioned and described a problem with the way a specific provision was transposed,
transposition performance was coded as 0. For example, one provision relates to the
partial reversal of the burden of proof, which states that “it shall be for the respondent to
prove that there has been no breach of the principle of equal treatment” and not for the
Based on the implementation reports, however, the Dutch legislation did not
apply the shift in the burden of proof in cases of victimization by the employer. In the
extreme case, Italy was criticized for failing to insert the provision on the shift of the
burden of proof in its national legislation. We coded member-state compliance with EU
provisions for both 2004 and 2006.
Independent variables: measurement
For the operationalization of goodness-of-fit we used an approach similar to Falkner
et al. (2005), who measured total misfit on an ordinal scale with three categories: ‘low’,
‘medium’ and ‘high’ misfit. In this study, the scale is reversed so that a higher value
refers to a better technical compatibility. The goodness-of-fit scale was constructed on
the basis of country reports for each of the 15 member states on anti-discrimination
legislation after the adoption of the Framework Equality Directive and before the
transposition process had started.
Based on the expert evaluations from the reports, we coded the level of fit as ‘low’ if
no policy or practice existed in a member state corresponding to a provision of the
Framework Equality Directive or the legal interpretation of the existing policy is
completely contradictory to the objective of the provision. The level of technical fit was
evaluated as ‘medium’ if one or more of the following conditions applied:
(1) Some practice exists, but no general policy;
(2) Some policy exists, but no general practice;
(3) National policies/practices have limited scope or should be reviewed to see
whether they meet the requirements of the provisions.
Technical fit was coded as ‘high’ if the country experts emphasized that few and not
substantive legal changes were necessary to fully comply with a provision. Based on the
data from the country reports, we were able to extract information on the goodness-of-fit
for 16 out of the 27 provisions (240 cases)
The measure on discretion employed in this study is based on previous research that
defines discretion as the number of provisions that grant powers to member states in the
implementation process relative to the total number of provisions in a directive (Epstein
and O’Halloran, 1999, Franchino, 2004; Thomson et al., 2007). In the present study,
however, discretion is measured at the provision level. Therefore, a distinction was made
between provisions that grant discretion to member states (coded as 1) and those that do
not (coded as 0) (Thomson, 2009). A provision grants discretionary powers if it allows
states to choose between different alternatives in the transposition process
. For example,
while the directive requires that member states should introduce legislation against age
discrimination, Art. 6 of the directive allows for differences in treatment under conditions
such as “fixing for occupational social security schemes of ages for admission or
entitlement to retirement or invalidity benefits” (Directive 2000/78/EC). We excluded
provisions that refer to delegating powers to the Commission. Thus, our measure of
discretion is based only on provisions relevant to the member states
Infringement procedures initiated by the Commission are measured at the member-
state level, since the reasons for their instigation related to the whole directive rather than
specific provisions (either non-notification or limited transposition). This variable is
coded as 1 if the Commission opened an infringement procedure against a member state.
Since we want to study the effect of infringements on the likelihood of compliance after
2004, we focused only on those infringement cases that were closed after 1
Conflict in the Council is measured at the provision level. Data on this independent
variable are taken from the “Decision-making in the European Union” (DEU) dataset,
which contains information on the policy positions of member states on the issues in the
proposal for the Framework Equality Directive that were debated during the Council
meetings (Thomson and Stokman, 2003; Thomson et al., 2006). One additional report
provided information on a number of issues that were debated before the Commission
designed the proposal. We closely read the informants’ reports to match each
controversial issue that caused debates during and before the Council meetings with the
respective provision in the directive
. A provision was coded as 1 if there was conflict in
policy preferences between the member state representatives.
Controls: corporatism and government position
We controlled for member-state-level characteristics related to the context of social
and labor law. For example, the level of corporatism is an important factor that could
affect member-state compliance with EU provisions in the context of employment
relations (Falkner et al., 2005). In more corporatist systems the social partners are
strongly embedded in national policy-making and could actively participate in the
transposition process. It is generally expected that such corporatist arrangements can lead
to more delays in the transposition process, since they increase the number of actors who
can veto a policy proposal (Thomson, 2009a; Falkner et al., 2005). However, corporatist
relations are also assumed to be conducive to the creation of policy networks with stable
coordination of government and interest groups, which improve compliance (Lampinen
and Uusikyla, 1998). The present study uses Siaroff’s (1999) quantitative index of states’
degrees of corporatism. The higher the values of the index, the higher the level of
integration of employers’ and employees’ organizations into the policy-making process
within a country.
Another important factor is the position of governments on the left/right dimension.
The political ideology of a government could have an effect on transposition success, as
it is generally expected that Center-Left governments are more likely to support
employment rights policies, while Center-Right governments represent the interests of
employers’ organizations and business associations. Given that the provisions of the
Framework Equality Directive increase employees’ protection against dismissal and
discrimination, it is plausible that Center-Left governments identify with the objectives of
the directive more than Center-Right governments, which leads to better transposition
performance of the former (Treib, 2003).
The control variable left/right position of the member state government is computed
on the basis of data from the “Party Policy in Modern Democracies” project (Benoit and
Laver, 2006). For coalition governments the positions are calculated by weighting the
score for each party participating in government by its relative share of votes and adding
these scores to form the average government position.
Description of member states’ transposition of Directive 2000/78/EC
Table 1 presents the distribution of the dependent variables employed in the analysis.
It shows that member states did experience problems in incorporating the EU provisions
into national legislation in 2004, when almost half of the cases were not transposed
correctly. Little progress was made by the end of 2006 as one-third of the transposition
cases were still incorrect.
Fig. 1 provides a clearer picture of the differences in transposition success across
different member states for 27 provisions of the Framework Equality Directive. It shows
the number of provisions that a member state had failed to comply with by the end of
2004 (dark) and 2006 (light-grey).
We observe some remarkable differences between standard measures of compliance
such as transposition delays and infringement procedures, and our measure: transposition
success with separate provisions
. First, while the UK, France and Italy did not
experience any delay with respect to the initiation of the transposition process, they failed
to comply with more than half of the provisions of the Framework Equality Directive.
Second, Commission documents reported that member states had transposed all
provisions by the end of 2006 (except for Sweden, which lacked legislation on the age
provisions). However, based on Fig. 1, we observe that correct transposition of all
provisions of the directive was not achieved by any of the member states, even if they all
had notified the Commission of transposition measures.
Thus, Fig.1 clearly illustrates the advantages of using direct measures of compliance
with separate provisions of a specific directive. It also lends support to the findings of
carefully documented case studies on the difficulties experienced by national legislators
when implementing the Framework Equality Directive (Bell, 2008).
Because the dependent variables on compliance are binary, we employ a binomial
logistic regression analysis. In addition, we apply a cross-classified design to account for
the dependence in the observations at the level of the member state and the provision
level. Cross-classified design is appropriate for the analysis in the present study, because
(1) different member states are transposing the same provisions and (2) different
provisions are being transposed by the same member states. This implies that we should
estimate the variation in member states’ compliance created by the crossing of two
random factors (provisions and member states) (Raudenbush and Bryk, 2002). Thus, we
improve on most existing literature, which hardly controls for cross-classified design of
Table 2 reports the results from the null model that tests for random effects both at the
provision-level and the member-state level (Rabe-Hasketh and Skrondal, 2008). The
random-effect parameters show significant residual standard deviation on the dependent
variables both between provisions and between member states. We observe more
variation in the likelihood of compliance across different provisions than across member
states. This observation supports our claim that differences exist not only in the
transposition across member states, but also in the transposition of different provisions
within a directive.
Table 3 presents the results from the analysis that tests our hypotheses on the
likelihood of member states’ compliance with the provisions of the Framework Equality
Directive for both 2004 and 2006. Models 1 and 2 test the effects on the two different
dependent variables, while Models 1b and 2b include the interaction effect to the
. The coefficients indicate the average change in the probability of correct
transposition of a provision due to changes in the values of a given independent variable.
A positive sign of the coefficients indicates that higher values of the independent
variables increase the probability of correct transposition of a provision, while a negative
sign shows that the probability of correct transposition decreases. The coefficients,
however, do not indicate the precise size of the effect for a change in a given variable x
since the change in the probability due to an increase in the level of this variable depends
on the levels of all factors in the analysis (Long, 1997).
We observe that the level of technical fit has a significant effect on the likelihood of
correct transposition for both 2004 and 2006 (Model 1a and 2a). The categories ‘low’ and
‘medium’ fit are compared to the reference category ‘high fit’. As predicted, the
likelihood of transposition success is significantly lower for cases with low and medium
fit relative to high levels of technical compatibility with the provisions of the directive.
To assess the magnitude of the effects, we transformed the coefficients into odds ratios.
Odds ratios indicate the factor change in the odds of compliance (relative to failure to
comply) due to changes in the values of an independent variable, holding all other
. In the present study, cases with low and medium fit had respectively
0.19 and 0.46 smaller odds of compliance than cases with high fit in 2004. At the end of
2006, the odds of transposition success were 0.24 (low fit) and 0.33 (medium fit) relative
to high fit.
The likelihood of successful transposition by member states depends also on whether
a provision grants discretionary powers to national policy-makers or not. The odds of
compliance in 2004 and 2006 are respectively 6 and 3 times greater for those provisions
that grant discretion than the provisions that do not grant discretion to member states.
Thus, member states are much more likely to correctly transpose provisions that give
implementers opportunities to choose from than provisions that constrain the choice of
implementers. This result is highly significant for both 2004 and 2006 and, thus, it is
consistent with our expectations.
Models 1b and 2b test the prediction on the interaction effect between the level of
technical fit and discretion (Hypothesis 2b). The reference category is the interaction
between ‘high fit’ and discretion. The results on both dependent variables show
significant differences between the marginal effects of ‘medium’ and ‘high’ levels of fit
on compliance when provisions do not grant any discretion to member states. We find
significant differences between low and high levels of fit when discretion is zero only on
the likelihood of compliance before the end of 2004. The coefficients in Table 3,
however, do not give information on the marginal effects of the different levels of
technical fit, when discretion is granted.
To provide a full test to the hypothesis, we computed the marginal effects and the
standard errors of low and medium levels of fit relative to high fit for discretion equal to
one. The marginal effects show that the odds of compliance for low levels of fit decrease
from 0.24 to 0.10 for 2004 and from 0.32 to 0.07 for 2006 as the discretion moves on the
scale from zero to one. This finding contradicts our hypothesis, because it suggests that
granting discretion to national authorities additionally impedes member states’
transposition success when there is a very low compatibility between national legal
structures and EU requirements. This effect is significant for compliance in 2006. When
we compare the marginal effects of medium and high fit on compliance for different
values of discretion, we arrive at the opposite conclusion. More precisely, the odds of
compliance for medium fit in 2004 and 2006 are respectively 0.37 and 0.25, when
discretion is zero, and 0.81 and 0.62, when discretion is one. This suggests that granting
discretion to national authorities increases the odds of compliance when there is medium
technical fit. The difference between the marginal effects for medium and high levels of
fit is no longer significant when discretion is granted, which is in congruence with our
One possible interpretation of our findings is that very low levels of technical and
legal compatibility are associated with lack of vital knowledge about the consequences of
implementing a particular provision. Granting discretion implies that member states have
different transposition alternatives at their disposal and some knowledge is necessary for
a national authority to be able to select an appropriate transposition measure. As a result,
if a member state must implement a provision that is completely missing in any of it
national legislation, discretion would create ambiguity about which transposition
alternative best suits the particular national legal context (Thomson et al., 2007).
To test the effect of the costs of persistent non-compliance by member states, we also
incorporated the effect of infringement procedures in the analyses on compliance by the
end of 2006. Based on the results from Table 3, the instigation of infringement
procedures by the Commission increased the probability that member states would
correctly incorporate provisions of the directive before 2007 (Model 2a and 2b). The odds
of compliance for member states, against which the Commission started formal
procedures, are three times greater than for member states, which were not confronted
with the threat of litigation. Infringement procedures seem to provide an incentive to
those member states with alleged non-compliant behavior to take measures to comply
with the provisions of the Framework Equality Directive.
Contrary to our expectations, we do not find significant differences in the
transposition of provisions that were debated in the Council and non-controversial
provisions. The non-significant coefficient could be partly due to the limited variation of
the variable, since only four of the provisions that were included in the final analysis
were controversial. Thus, in the context of the Framework Equality Directive we do not
find support for the hypothesis on the effect of the divergence of member states’ policy
preferences on the likelihood of member states’ compliance with EU provisions.
Table 3 also presents the estimates on the effects of the control variables included in
the analysis. The results show that more corporatist countries were less likely to
experience compliance problems in 2004. A change of one point in the scale of
corporatism leads to approx. 1.5 greater odds of successful transposition. However, the
effect of corporatism is not significant in the analysis on member states’ transposition
success at the end of 2006. Government position on the left/right dimension is not
significant in any of the models.
The present study was designed to test hypotheses on member-state compliance with
different provisions of an EU law. Thus, we relaxed the assumption that member states
only transpose directives as a whole. Instead, they might comply with certain aspects of a
directive while failing to comply with others. To explain provision-level variation in
compliance we formulated hypotheses on member states’ adaptation costs and perceived
consequences of non-compliance. We also focused on the correctness of transposition
measures adopted by national policy-makers instead of the timeliness of transposition.
Finally, we analyzed the transposition of the Framework Equality Directive at two
different periods of time. This directive provided an interesting case to analyze provision-
level differences due its high level of controversy and numerous policy aspects. It also
shows member-state compliance with provisions on a specific EU legislation that touched
upon a salient and politically sensitive topic.
Our analyses reveal a number of interesting findings. First, we observed great
variation in compliance both across different member states and different provisions of
the Framework Equality Directive. Our measure of compliance indicated poor
transposition of the provisions of the directive years after the deadline had passed, despite
the fact that all member states had communicated transposition measures to the
Commission. This result is in congruence with recent discussions on the methodological
problems related to commonly used data from the Commission databases (Hartlapp and
Falkner, 2009). It also supports the findings from case studies on the problems
encountered by member states in the implementation of the Framework Equality
Directive (Bell, 2008). While the majority of case studies focus on the problems
encountered by the ‘new’ member states (Dimitrova and Rhinard, 2005; Falkner and
Treib, 2008), our analyses showed that the 15 ‘old’ member states did not perform better
in transposing the provisions of the directive. We improve on existing case studies by
systematically analyzing the variation that exists across all relevant provisions instead of
focusing on only few important aspects of a directive.
Second, our findings reveal that domestic costs of adaptation to the EU requirements
play an important role in defining the choice and opportunities of national policy makers
to comply with EU provisions. Thus, member states are more likely to comply with
provisions that require only marginal changes to existing legal documents (Börzel, 2003).
Furthermore, we find evidence that provisions granting discretion to member states are
more likely to be successfully transposed than provisions that constrain the actions of
national policy-makers. This finding supports the theory that discretion facilitates
compliance (Knill, 2001).
Our expectation that discretion downplays the difference between higher and lower
levels of fit on member-state transposition success is only partially supported. We find
that discretion facilitates member states’ transposition success for medium levels, but not
for low levels of technical fit. Given the small number of observations in the present
study, future work should shed more light on the interplay between different levels of fit
and discretion on compliance with EU provisions.
Our findings show that the Commission behavior affects member states’ transposition
success. More precisely, member states that received a formal warning by the
Commission achieved a better transposition record by the end of 2006 than member
states that were not sanctioned. This finding supports theories on supranational
enforcement of EU law that the infringement procedure forces member states to correct
mistakes in the transposition process out of fear of litigation and subsequent sanctions
(Tallberg, 2003). By contrast, conflict in the Council did not have a significant impact on
member states’ transposition success.
With a view to the goals of the present study, our findings should be put into a proper
perspective. We studied member states’ compliance with different provisions of only one
directive. Ideally, it would have been better to test our predictions on different provisions
in several directives. However, given the nature of the dependent variable (transposition
success), collecting such data is a complicated and time-consuming process. On the
flipside, this research design allows us to focus on variation in compliance due to
provision-level factors, without confounding effects from directive-level factors. As a
result, our study should be seen as a first attempt to uncover the mechanisms that drive
compliance with separate provisions and the extent to which general implementation
theories can be applied to a lower level of analysis. This point has implications for future
research, since many of the common directive-level factors on compliance have no
equivalent at the provision level. Further work should shed some light on whether
alternative operationalizations of goodness-of-fit and discretion could better explain the
transposition success of member states with separate provisions. For instance, our
measure of fit excludes member state’s normative costs for implementing anti-
discrimination policies, which are very relevant for at least some of the provisions of the
Framework Equality Directive (Dimitrova and Rhinard, 2005).
Future work should, also, incorporate more precise information about member states’
policy preferences on specific provisions. This could explain the non-significant effect of
preference-based explanations in the literature once these preferences are aggregated to
the level of the directive (Thomson, 2009b).
More general indicators for adaptation costs can be derived from different
conceptualizations of ‘goodness of fit’. A distinction is made between institutional or
structural fit (Knill and Lenschow, 1998) and ‘policy fit’. Another distinction concerns
the difference between formal and informal policy fit (Falkner et al., 2005).
The reports were published by the European Monitoring Centre on Racism and
Xenophobia (EUMC) in 2002 as part of the project ‘Implementing European Anti-
The reports lacked information about the level of fit on the protection of disabled
persons or anti-discrimination practices on age and other related provisions.
It is important to note that provisions that grant discretion are not necessarily associated
with high goodness-of-fit. An example is the provision on positive action to favor
disadvantaged groups. In this case, the level of fit is measured as the extent to which
member states had existing practices on positive action. Greece and Finland, for instance
lacked tradition in positive action. The provision, however, gave a lot of flexibility to
member states in the extent to which they can compensate disadvantaged groups.
Since provision-level measures of discretion do not currently exist, we follow the logic
of existing directive-level measures. A binary variable should be sufficient to capture the
differences between more and less restrictive provisions.
For example, one particular point of debate concerned what approach should be adopted
to tackle discrimination against disabled persons: “a generic approach” vs. “reasonable
adjustment”. This controversial issue was matched with the provision on “reasonable
accommodation of disabled people” in the directive.
Note that transposition progress refers to the extent to which member states fully
complied with the provisions of the directive. Adjustments that partially improved the
level of compliance with an EU provision are left out of the analysis, if the informants
stated that transposition problems still existed.
We also pooled the compliance data for both time periods into a single regression to
check for the robustness of the results. The alternative analysis produced very similar
estimates to the ones reported in Table 3.
Thus, given that β
is the estimated coefficient for the effect of a variable x from the
logistic regression model, for a change of δ in the values of x, the odds are expected to
change by a factor of exp(β
X δ), holding all other variables constant (Long, 1997)
Bailey, I. (2002) ‘National adaptation to European integration: institutional vetoes and
goodness-of-fit’, Journal of European Public Policy 9(5): 791-811.
Bell, M. (2008) ‘The Implementation of European Anti-Discrimination Directives:
Converging towards a Common Model?’, The Political Quarterly 79(1): 36-44.
Benoit, K. and Laver, M. (2006) Party Policy in Modern Democracies. London:
Börzel, T. A. (2001) ‘Non-compliance in the European Union: pathology or statistical
artefact’, Journal of European Public Policy 8: 803-24.
Börzel, T. A. (2003) Environmental Leaders and Laggards in Europe: Why there is (not)
a ‘Southern Problem’. Aldershot: Ashgate.
Casado Asensio, J. (2008) ‘Transposition Beyond Policy Legacies and Veto Players?
The Case of Anti-Discrimination Policy in Austria’, Minority Policies 5: 153-74.
Dimitrova, A. and Rhinard, M. (2005) ‘The Power of Norms in the Transposition of EU
Directives’, European Integration Online Papers (EIoP) 16(9)
Epstein, D. and O’Halloran, S. (1999) Delegating Powers. A Transaction Costs Politics
Approach to Policy Making Under Separate Powers. Cambridge: Cambridge
Falkner, G., Hartlapp, M., Leiber, S. and Treib, O. (2005) Complying with Europe: EU
Harmonization and Soft Law in the Member States. Cambridge: Cambridge
Falkner, G. and Treib, O. (2008) ‘Three Worlds of Compliance or Four? The EU-15
Compared to New Member States’, Journal of Common Market Studies 46(2):
Franchino, F. (2004) ‘Delegating Powers in the European Community’, British Journal
of Political Science 34(2): 269- 93.
Franchino, F. (2007) The Powers of the Union: Delegation in the EU. Cambridge:
Cambridge University Press.
Hartlapp, M. and Falkner, G. (2009) ‘Problems of Operationalization and Data in EU
Compliance Research’, European Union Politics 10(2): 281-304.
Haverland, M. (2000) ‘National Adaptation to European Integration: The Importance of
Institutional Veto Points’, Journal of Public Policy 20(1): 83-103.
Héritier, A. (1996) ‘The Accommodation of Diversity in European Policy Outcomes:
Regulatory Policy as a Patchwork’, Journal of European Public Policy 20: 83-
Héritier, A., Kerwer, D., Knill, C., Lehmkuhl, D., Teutsch, M., and Douillet, A-C. eds.
(2001). Differential Europe: The European Union Impact on National
Policymaking. Lanham, MD: Rowman and Littlefield.
Hill, M. and Hupe, P. (2002) Implementing Public Policy: Governance in Theory and in
Practice. London / Thousand Oaks / New Delhi: Sage.
Kaeding, M. (2008) ‘Lost in Translation or Full Stream Ahead: The Transposition of EU
Transport Directives across Member States’, European Union Politics 9: 115 –
Knill, C. and Lenschow, A. (1998) ‘Coping with Europe: The Impact of British and
German administrations on the implementation of EU environmental policy’,
Journal of European Public Policy 5: 595 – 615.
Knill, C. (2001) The Europeanisation of National Administrations. Patterns of
Institutional Change and Persistence. Cambridge: Cambridge University Press.
König, T. and Luetgert, B. (2009) ‘Troubles with Transposition? Explaining Trends in
Member-State Notification and Delayed Transposition of EU Directives’, British
Journal of Political Science 39: 163- 94.
Lampinen, P. and Uusikylä, P. (1998) ‘Implementation Deficit – Why Member States Do
Not Comply with EU Directives?’, Scandinavian Political Studies 21: 231 – 51.
Long, J.S. (1997) Regression Models for Categorical and Limited Dependent Variables,
Vol. 7 of Advanced Quantitative Techniques in the Social Sciences. Thousand
Oaks, CA: Sage Publications.
Majone, G. (2001) ‘Two Logics of Delegation. Agency and Fiduciary Relations in EU
Governance’, European Union Politics 2: 103-22.
Mastenbroek, E. (2003) ‘Surviving the Deadline: The Transposition of EU Directives in
the Netherlands’, European Union Politics 4(4): 371-95
Mastenbroek, E. (2005) ‘EU Compliance: Still a ‘Black Hole’?’, Journal of European
Public Policy 12(6): 1103 – 20.
Mastenbroek, E. and Kaeding, M. (2006) “Europeanization Beyond the Goodness of Fit:
Domestic Politics in the Forefront”, Comparative European Politics 4(4): 331-54.
Mbaye, H. A. D. (2001) ‘Why National States Comply with Supranational Law:
Explaining Implementation Infringements in the European Union, 1972-1993’,
European Union Politics 2: 259-81.
McCubbins, M. D., Noll, R. G. and Weingast, B. R. (1989) ‘Structure and Process as
Solution to the Politician’s Principal Agent Problem’, Virginia Law Review 75:
O’Toole L. J. Jr. (2000) ‘Research on Policy Implementation: Assessment and
Prospects’; Journal of Public Administration Research and Theory 10: 263-88.
Perkins, R. and Neumayer, E. (2007) ‘Do Membership Benefits Buy Regulatory
Compliance? An Empirical Analysis of EU Directives 1978–99’, European
Union Politics 8(2): 180–206.
Rabe-Hesketh, S. and Skrondal, A. (2008). Multilevel and Longitudinal Modeling Using
Edition). College Station, TX: Stata Press.
Raudenbush, S. W. and Bryk, A. S. (2002). Hierarchical linear models: Applications
and data analysis methods. 2
edition. Newbury Park, CA: Sage.
Siaroff, A. (1999) ‘Corporatism in 24 Industrial Democracies: Meaning and
Measurement’, European Journal of Political Research 36: 175-205.
Steunenberg, B. (2006) ‘Turning Swift Policymaking into Deadlock and Delay: National
Policy Coordination and the Transposition of EU Directives’, European Union
Politics 7: 293–319.
Steunenberg, B. and Toshkov, D. (2009) ‘Comparing transposition in the 27 member
states of the EU: the impact of discretion and legal fit’, Journal of European
Public Policy 16(7): 951 – 70.
Tallberg, J. (2003) ‘Paths to Compliance: Enforcement, Management, and the European
Union’, International Organization 56: 609 – 43.
Thomson, R. and Stokman, F. N. (2003) Decision making in the European Union. [machine
readable data set]. Groningen: ICS.
Thomson, R. (2009a) ‘Same Effects in Different Worlds: The Transposition of EU
Directives’, Journal of European Public Policy 16(1): 1-18.
Thomson, R. (2009b) ‘Opposition through the Back Door in the Transposition of EU
Directives’, Paper presented at the Fifth General ECPR Conference in Potsdam
11-13 September, 2009
Thomson, R., Stokman, F. N., Achen, C. and König, T (eds.) (2006) The European
Union Decides. Cambridge: Cambridge University Press
Thomson, R., Torenvlied, R. and Arrequi, J. (2007) ‘The Paradox of Compliance:
Infringements and Delays in Transposing European Union Directives’, British
Journal of Political Science 37: 685 – 709.
Torenvlied, R. (2000) Political decisions and agency performances. Dordrecht / Boston:
Kluwer Academic Publishers.
Treib, O. (2003) ‘EU Governance, Misfit, and the Partisan Logic of Domestic
Adaptation: An Actor-Centered Perspective on the Transposition of EU
Directives’, Paper presented at the EUSA 8
International Biennial Conferences,
Nashville, Tennessee, 27-29 March, 2004.
Zhelyazkova, A. and Torenvlied, R. (2009) ‘The Time-Dependent Effect of Conflict in
the Council on Delays in the Transposition of EU Directives’, European Union
Politics 10(1): 35 – 62.
Tables and Figures:
Table 1: Member-state compliance with the provisions of Directive 2000/78/EC
Compliance 2004 Compliance 2006
None or incorrect transposition 174 (45%) 136 (34%)
Correct transposition 212 (55%) 261 (66%)
386 (100%) 397 (100%)
Fig. 1: Member state’s noncompliance with separate provisions at the end of 2004 and 2006
0 5 10 15 20
uk el de lu fr it es nl ir dk at se pt be fi
problematic provisions (2004) problematic provisions (2006)
Table 2: Cross-classified logistic regression: null model
.350(.451) .902 (.336)
S.d. (Provision-level) 1.800(.344) 1.321(.258)
S.d. (member-state-level) .959(.248)
Log likelihood -213.674 -224.014
The actual number of cases differs from the maximum possible number of 405 (15 X 27), because we
could not identify the transposition performance of member states in all cases. In addition, there are more
missing cases on the dependent variable for 2004 than for 2006.
Table 3: Cross-classified logistic regression on the likelihood of compliance with EU provisions
Compliance 2004 Compliance 2006
Model 1a Model 1b Model 2a Model 2b
Low fit -1.661
Medium fit -.779
Infringement procedure ----- ----- 1.097
Conflict in the Council
-1.267 (.921) -1.391 (1.239) -.998 (.768) -1.019 (.736)
Low fit * Discretion -----
Medium fit * Discretion -----
(.173) .204 (.145) .223 (.151)
Left/Right position .025(.073) .023(.073) .058(.060) .058(.062)
S.d. (Provision-level) 1.442(.382) 1.410(.394) 1.164(.310) 1.094(.311)
S.d. (member-state-level) .431(.284)
Log likelihood -120.271 -119.426 -126.851 -124.934
N 229 234
High level of fit and High fit * Discretion are the reference categories; Random effects estimated for
16 provisions and 15 member states; *** = p < .01, ** = p < .05, * = p < .10