Signalling ‘compliance’: The link between notified EU directive implementation and infringement cases

Introduction

The success of European integration depends on the extent to which the decisions adopted by the European Union (EU) are implemented by its member states. It is then not surprising that research on national compliance with EU legislation is abundant. Many studies equate member states’ implementation performance with the speed or timeliness of self-reported legal measures adopted at the national level to meet specific EU requirements (see, for example, Berglund et al., 2006; Haverland et al., 2011; König and Luetgert, 2009; Luetgert and Dannwolf, 2009; Mastenbroek, 2003; Thomson et al., 2007). In particular, the member states are expected to incorporate the EU directives into their national legal frameworks, meeting strict deadlines but with instruments of their choice. Data on the implementation performance are generally based on the extent to which member states notify the EU Commission of transposition measures in a timely manner. There is, however, little knowledge about the quality of the notified measures and the extent to which they are even relevant for the implementation of the EU directive at hand. This is problematic because reporting measures before the deadline may not necessarily mean that national authorities have fully implemented the EU requirements and compliance problems may persist. The question that arises then is under what conditions self-reported measures are associated with compliance problems.

To address this question, the article focuses on a particular type of implementing instruments: domestic legal measures that a member state reported to the Commission for the implementation of a given directive, but which entered into force before the directive was even adopted by the EU. Member states could have different incentives to report pre-existing measures. On the one hand, the established view is that pre-existing measures reflect member states’ ability to ‘upload’ their preferences during the EU decision-making stage (Kaeding, 2006; Linos, 2007; Mastenbroek, 2003). In a similar way, early measures may represent member states’ ‘policy fit’ with an EU directive (Thomson et al., 2007). Based on these assumptions, pre-existing measures are expected to increase member states’ compliance with EU law. On the other hand, research so far has not considered that pre-existing measures are also likely to fulfill only partially the compliance requirements and member states may report them as a means to win time before an infringement procedure escalates to later stages that involve tangible sanctions. Specifically, instead of missing the transposition deadline a member state could notify pre-existing measures as a strategy to disguise the fact that it has not yet (fully) transposed an EU directive and buy time to complete the transposition process at a later point in time. Reporting pre-existing measures could also reflect the unwillingness or inability of national authorities to change their current legislation to fully conform to the EU standards. These assumptions lead to the contrasting prediction that instead of increasing the probability of compliance, notifying pre-existing measures is linked to actual implementation problems. In this article, we try to disentangle these contrasting mechanisms by taking into consideration the level of fit of pre-existing measures relative to the EU requirements, member states’ subsequent implementation measures and their incentives to fully comply given the expected reputational costs for not doing so.

We rely on a new data set on EU directives that were adopted and had to be transposed in the period between May 2004 and May 2010. The results suggest that, contrary to common expectations, notifying pre-existing measures significantly increases the likelihood of implementation problems and, hence, infringement cases. We find that this is particularly the case when those measures are subsequently adjusted only after the transposition deadline has passed. We further show that the positive relationship between early measures and non-compliance detected by the Commission is conditional on the policy fit of pre-existing measures and it ceases to exist whenever transparency in the implementation process is explicitly required. These findings bear important implications for the usefulness of self-reported transposition measures as indicators for compliance with the EU directives, especially because many EU scholars focus on the earliest notified legal instrument to estimate member states’ actual behaviour (Berglund et al., 2006; Luetgert and Dannwolf, 2009; Thomson et al., 2007; Zhelyazkova and Torenvlied, 2009). More substantively, we show that the EU can induce compliance by demanding a more transparent implementation process in the member states.

Self-notified transposition of EU law and detected non-compliance

Self-reported implementation measures are an important tool for signalling compliance with international agreements. In particular, self-reported measures decrease information asymmetries about states’ activities to comply with their international obligations. However, without any external evaluation, it is not possible to know whether a government reveals truthful information or not. This makes self-reporting informative only if a member state has incentives to comply with its international obligations and to report correct information about its implementation activities. Instead, if the national government faces high adjustment or political costs of complying, self-reporting may not be a useful indicator of a state’s compliance record and national implementers could even exaggerate their implementing activities. In such cases, self-reporting could be ‘cheap talk’ (Crawford and Sobel, 1982; Gneezy, 2005), where the act of sending a signal incurs small, or no costs on the sender and the signal could be either truthful or not. In other words, a government could declare that it has complied with an international agreement, while this is not the case in reality.

There could be different reasons why states report non-compliant implementing measures. On the one hand, states could send such measures because they are unable in the short run to muster the necessary resources to fully comply and need time to fulfill their obligations. By strategically sending partially relevant measures, national governments at least signal that they have started the implementation process and, thus, could gain time to acquire sufficient resources to fulfill their international obligations. On the other hand, governments’ implementation activities may be driven by their general willingness to conform to international requirements. In this case, the extent to which a state government strategically engages in ‘cheap talk’ without fully complying depends on the expected reputational losses it would incur if compliance problems were detected. Provided that any negative consequences may take time to materialize, reporting only partially- or non-compliant implementing measures could serve as a time-winning strategy for short-sighted democratically elected governments to avoid the blame for non-compliance or shift it to their successors. Even if non-compliance becomes internationally known, a wayward government may still expect low reputational losses for agreements that its international partners do not consider very important.

In this article, we study the link between EU member states’ self-reported activities to implement European legislation and detected non-compliance by the European Commission. The EU is a suitable case to study a state’s response to its international obligations because national and sub-national actors are obliged to incorporate, enforce and apply the EU adopted decisions within national settings (Börzel et al., 2005). In particular, EU directives stipulate the objectives that must be attained by the member states, but give national authorities a certain degree of discretion in choosing the appropriate means to attain these objectives. Directives are, thus, not directly enforceable but need to be formally incorporated into member states’ legislation: a process that is referred to as transposition. As transposition is executed at the national level, the type of transposition measures and the extent to which they actually meet the EU requirements are not commonly known by other member states and the Community as a whole. To decrease information asymmetries, member states are to notify the Commission before a specified deadline of all relevant measures that implement a directive. If a member state fails to meet the deadline, the delay is interpreted as a deviation from the EU law (Thomson et al., 2007; Zhelyazkova and Torenvlied, 2009), in response to which the Commission opens an infringement procedure consisting of three formal stages: a letter of formal notice, a reasoned opinion, and, ultimately, a referral to the European Court of Justice (ECJ). With each stage of the infringement procedure, the Commission increases the pressure on member states to comply. Provided that violations persist, as a measure of last resort the Commission can propose financial penalties to be imposed on member states. However, it could take years until monitory sanctions materialize and earlier stages of the infringement procedure are considered a weak form of enforcement (Tallberg, 2002).

It is important to note that the use of infringement cases as an indicator for non-compliance is generally criticized by scholars (Börzel, 2001; Hartlapp and Falkner, 2009; König and Luetgert, 2009; König and Mäder, 2014; Thomson et al., 2007). This is because the Commission has limited capacity to detect all compliance problems in each member state and it could even choose not to pursue cases of non-conformity that are in congruence with its policy objectives or could impose high litigation costs. Establishing non-compliance is particularly problematic in cases of wrong transposition and problems of practical implementation, where the Commission lacks resources to monitor compliance. For these reasons, the Commission has focused its monitoring activities on checking the timing and completeness of member states’ self-notified measures, where problems of compliance are much easier to detect. In the Single Market Scoreboard, the Commission often indicates that it is in the progress of checking the relevance and the completeness of the notified measures by a member state. Being considered ‘the guardian of the Treaties’, the Commission is also likely to at least note cases of non-compliance, even if it subsequently decides not to pursue litigation against non-compliant member states. As a result, infringement proceedings are likely to be a suitable indicator of non-compliance at least in those cases related to the timing of the transposition process.

In general, there is little discussion of the extent to which self-reported measures fully correspond to the requirements specified in EU directives. This is likely due to the high number of measures that member states notify to the Commission, which creates difficulty for scholars to analyze the relevance of each individual measure. In this study, we focus on a particular type of self-reported measures: notified measures pre-dating not only the implementation deadline, but also the official date of adoption of a directive at the EU level. By reporting pre-existing measures, a government claims that it has already (at least partially) implemented the respective EU directive even before the directive was adopted. Thus, sending pre-existing measures constitutes ‘cheap talk’, where the act of sending such measures does not incur costs for the governments (since domestic legislation was already in force) and it does not tell us whether the reported information is compliant or not.

While pre-existing measures could signal a member state’s readiness to meet the EU requirements, or indeed the fit of their existing legislation to an EU directive, notifying such early measures could be associated with compliance problems. Under certain circumstances, the governments may notify only partially relevant or even incompatible measures, leading to the Commission opening infringement cases against member states for incomplete directive implementation. Below we investigate the conditions under which we can expect that.

Research design

To test our hypotheses, we have gathered data on all EU directives that were adopted after May 2004 (when 10 countries acceded to the EU) and had their transposition deadline before May 2010. These selection criteria ensure that both the old and the new EU member states at that time participated in drafting the directives in our sample and, hence, had the chance to upload their preferences in the law-making process. Our data includes all the directives adopted by the Council of Ministers and the European Parliament (97 in total), but excludes Commission directives. The latter are rather technical, provide less room for member states’ implementation activities and are less politically important. We used the Legislative Observatory of the European Parliament to identify the sample (European Parliament, 2011). The observations in our dataset are directive–country dyads. ¹

For our dependent variable measure, we consulted the Annual Reports of the Application of Community Law published by the EU Commission for each of the years until 2011 (European Commission, 2011). These reports contain information about the infringement procedures that have been initiated against member states for non-timely or incomplete transposition, ² and thus provide data for our dependent variable. ³ The advantage of using the Commission reports is that they provide detailed information about all individual infringement cases (including letters of formal notice, reasoned opinions, referrals to the ECJ) sent to a member state regarding a particular EU directive. Letters of formal notice are usually used by the Commission to eliminate those cases that were issued due to uncertainty about the member states’ non-compliance. As reasoned opinions constitute the first formal stage at which the Commission establishes that a member state is in violation with an EU law, we use them to measure detected non-compliance and test our hypotheses. In this way, we reduce the potential that the Commission’s response to early measures could be the result of a mere suspicion and not necessarily related to non-compliance. ⁴

To measure our main independent variable, namely the notification of a pre-existing measure, we relied on the EurLex (2011) database, which contains information about all of the legal measures that member states notified to the European Commission as steps taken to transpose EU directives into their national legal frameworks. It is common for member states to report more than one measure. To test H1, we focus on the timing of the first notified measure and distinguish between: (1) measures that have already been in force before the date of adoption of a particular EU directive, which we call Pre-existing measures; (2) measures that have been published or enforced by national authorities after the respective directive’s adoption but before its transposition deadline (timely measures); and (3) measures that were published or enforced by national authorities only after the transposition deadline (Late measures). ⁵

To illustrate the frequency with which member states report pre-existing measures, Figure 1 shows the average probabilities of member states notifying the Commission of pre-existing measures regarding the transposition of a particular directive, which were computed using logistic regression models. We observe that the majority of new member states except Estonia, Malta and Cyprus have a significantly higher probability of reporting pre-existing measures than the old member states (with the exception of Finland and Sweden). It is important to note that this is not an indicator of higher non-compliance in these countries by itself. Yet, Figure 1 shows that reporting early measures is a very common practice for at least some of the EU member states, which makes the question about the relevance of these measures even more important. OPEN IN VIEWER

To evaluate H2a, or when such reporting results in an infringement case, we consider whether a member state notifies only pre-existing measures (Only pre-existing measures) or whether it subsequently reports other measures and, if so, whether the latter are then timely or only enforced after the transposition deadline has passed (Only pre-existing and late measures).

As argued above, the propensity of some member states to report pre-existing measures does not tell us to what extent these measures actually fit the goals of the directives. Yet, finding a policy fit measure for a large sample of directives and member states is challenging, as it is practically infeasible to compare the text of all notified early measures to the goals of the directives. Instead, following Steunenberg and Toshkov (2009), we rely on information about the nature and number of all subsequent measures notified by member states to gain an understanding about the relevance and the fit of the reported pre-existing measures and, thus, to test H2b. First, if a member state needed to notify many instead of only a few additional measures following its pre-existing national legislation, it is likely to have lower fitting early measures. Second, subsequently reported measures indicate a better fit if they primarily amend the pre-existing legislation rather than introduce entirely new legislation (Kaeding, 2006; Mastenbroek, 2003). ⁶

We combine these characteristics to construct an early measure Policy fit index of four categories. Thus, a member state is least likely to have well-fitting early measures if the respective national authorities report at a later stage many measures (more than three), the majority of which are new rather than amending (Low policy fit). At the other end of the scale, the notified early measures are likely to be highly fitting if a member state reports subsequently just a few pieces of legislation (up to three), most of which are amending rather than new (High policy fit). In between lie cases of Limited policy fit of pre-existing legislation (more than three, mostly amending subsequent measures) and Medium policy fit of pre-existing legislation (up to three, mostly new subsequent measures). Arguably, the measure for policy fit runs counter to a possible assumption that member states who want to conceal their non-compliance are likely to report only pre-existing measures. Instead, we assume that low-fitting pre-existing measures are associated with a high demand for subsequent implementing instruments, which would induce governments to report many new measures at a later stage. Furthermore, reporting many new timely measures does not necessarily mean that these are compliant. It does, however, indicate that there was a high necessity to implement many measures in addition to the pre-existing legal instruments. ⁷

Finally, to test H3 we need to account for the importance of directive transposition for other member states and/or the Commission and, hence, the reputational costs a national government would incur for not complying. For the purpose, we consider the explicit inclusion of a transparency requirement for the implementation process in the text of an EU directive. Specifically, we look at whether member states are required to provide concordance tables to the Commission with information about the transposition of each directive provision (coded as 1, otherwise 0). ⁸ From the total sample of directives (97), 23 directives included a transparency requirement.

In addition, the analysis includes a number of control variables to consider that member states may have different incentives to report early measures. Thus, we include a variable for member states’ voting weights based on the Nice Treaty to control for whether more powerful member states report early measures because they managed to upload their preferences during the decision-making stage. Moreover, the member state that held the presidency of the Council of Ministers when the directive was adopted could have steered the negotiations in favour of its own positions (coded as 1, otherwise 0).

Furthermore, the extent to which member states have incentives to send low-fitting early measures depends on their willingness and capacity to comply with the EU directives. As an indicator of a member states’ willingness, we use data on Governmental EU support from the Chapel Hill expert survey on party positions (Bakker et al., forthcoming). This measure is calculated by first identifying the cabinet parties at the time of a directive’s adoption using the database of Döring and Manow (2010) and then taking the average of these parties’ EU positions weighted by the parties’ relative share of parliamentary seats. As an indicator of state capacities, we employ a measure of Governmental effectiveness, which captures a number of governmental characteristics, such as the quality of the public and civil services and the degree of independence of civil servants from political pressure. This measure is based on the World Bank’s Governance indicators (Kaufmann et al., 2010) and is meant to control for the extent on which member states’ compliance records depend on their administrative capabilities to deal with the workload that transposition entails (Thomson, 2009; Toshkov, 2008). We took the average score of governmental effectiveness for the whole period allocated to a directive’s transposition. In addition, government incentives to report non-fitting early measures could increase if the government that initiated the transposition process stepped down before the transposition process was completed and the new government was pressured to meet the transposition deadline. Thus, we control for a Change in government to account for possible discontinuities in the national policy-making process. It could also be argued that the older 15 member states have more relevant pre-existing measures than the new entrants of the 2004 enlargement. Lastly, the analysis controls for the impact of directive-level characteristics on infringement cases. In particular, we control for the number of Recitals in a directive as an indicator for policy complexity. National authorities may report many implementing measures because they are confronted with complex directives instead of having low-fitting pre-existing legislation.

The online appendix provides descriptive statistics of all the variables used in the analysis.

Analysis

To test our hypotheses, we apply logistic regression because the issue of a reasoned opinion is a binary variable. Furthermore, we use a multilevel cross-classified design to account for the unobserved variance at the directive and the member state levels. This is necessary because transposition outcomes are simultaneously nested in two higher levels – member state and directive – which requires the inclusion of random effects for each of these levels in the statistical models. ⁹

Table 1 presents the results from the crossed-effects logistic regression analysis of the initiation of infringement proceedings against a member state for non-compliance with the transposition of an EU directive. Each model tests one of the four hypotheses regarding the effect of sending pre-existing measures on (detected) non-compliance by the Commission. The table presents the coefficients, which indicate the direction but not the size of the effect of a unit change in a given variable on the likelihood of receiving a reasoned opinion from the Commission (Long, 1997). To assess the magnitude of the effects, we transformed the coefficients into odds ratios (not reported), which we discuss in the text. The odds ratios indicate the factor change in the odds of a member state receiving a reasoned opinion with regards to the transposition of an EU directive due to a unit increase in the value of an independent variable, holding all other variables constant at their means. As we model the nested structure of the data, we also report the variance components at the levels of the directive and the member state.

OPEN IN VIEWER

Table 1.

Crossed effects multi-level logistic regression of the issue of a reasoned opinion.

	Model 1	Model 2	Model 3	Model 4
Pre-existing measure	0.793**
	(−0.229)
Late measure	1.221**	0.961**	1.194**
	(0.188)	(0.164)	(0.187)
Only pre-existing measures		−1.684
		(1.041)
Only pre-existing & late measures		0.671*
		(0.266)
Low policy fit			1.677**
			(0.288)
Limited policy fit			0.907**
			(0.335)
Medium policy fit			0.012
			(0.329)
High policy fit			0.180
			(0.389)
Policy fit index				−0.868**
				(0.154)
Transparency requirement	−1.062**	−1.106**	−1.024**	−3.789**
	(0.371)	(0.379)	(0.361)	(1.281)
Transparency req.*Policy fit				0.880*
				(0.411)
Voting weights in Council	0.015	0.015	0.015	0.004
	(0.016)	(0.016)	(0.016)	(0.037)
Presidency	0.300	0.311	0.292	0.259
	(0.317)	(0.316)	(0.318)	(0.950)
Governmental EU support	0.153	0.183	0.125	0.162
	(0.106)	(0.108)	(0.108)	(0.190)
Governmental effectiveness	−0.710	−0.674	−0.711*	0.543
	(0.363)	(0.365)	(0.358)	(0.922)
Change in government	0.505**	0.420**	0.527**	0.751*
	(0.151)	(0.152)	(0.152)	(0.299)
Recitals	0.030**	0.030**	0.026**	0.019
	(0.008)	(0.008)	(0.008)	(0.012)
Old member state	0.939*	0.721	0.938*	−0.487
	(0.412)	(0.407)	(0.409)	(0.960)
Random effects
σ 2 (directive level)	1.470	1.562	1.353	1.649
	(0.329)	(0.346)	(0.311)	(0.637)
σ 2 (member state level)	0.237	0.242	0.227	0.61
	(0.102)	(0.104)	(0.100)	(0.441)
Log-likelihood	−824.7	−825.3	−810.4	−240.0
Observations	2033	2033	2033	636

Model 1 in Table 1 tests the popular view that reporting pre-existing legislation (be it even only one pre-existing measure) reflects a state’s readiness to fully comply with an EU directive and, thus, decreases the likelihood that the Commission will issue a reasoned opinion (H1). We control for cases where a member state sent only late measures. Late transposition should automatically lead to infringement cases and including it in the reference category would artificially inflate the suggested negative effect of reporting pre-existing measures on infringement cases. In model 1, the reference category thus includes all other cases where a member state did not start the transposition process with a pre-existing measure but with a timely one (adopted after the directive adoption and before the transposition deadline).

In contrast to the expectation in H1, the results show that a member state is significantly more likely to receive a reasoned opinion from the Commission if it reports a pre-existing measure (i.e. a measure enforced before the adoption of a directive) than if it still meets the transposition deadline but starts the transposition process after the adoption of the directive. Reporting pre-existing measures increases the odds of receiving a reasoned opinion by 124 per cent. Rather than reflecting a member state’s ability to upload its domestic policies at the EU level, the finding is in congruence with the conjecture that not all pre-existing measures tend to meet the requirements of a directive.

In models 2 and 3, we test the conditions under which reporting pre-existing measures leads to infringement cases. First, we expected to find that reporting early measures without subsequently supplementing them with new domestic legislation before the transposition deadline is more likely to be associated with compliance problems (H2a). In model 2, we test this hypothesis, making a distinction between cases where a member state reported only pre-existing measures and cases where a member state did subsequently report additional measures but only after the transposition deadline (i.e. early measures were followed by late measures only). Thus, in this model the reference category is all cases where a member state reported at least one domestic measure that was enforced after the adoption of the EU directive but before the transposition deadline.

The results show that reporting only pre-existing measures does not significantly decrease the likelihood that a member state will get an infringement case. The lack of significance in the coefficient could be due to the very limited number of such cases in our data (less than 3 per cent). However, the negative sign suggests that in those cases, where a government reported only pre-existing measures, these measures were actually sufficient to convince the Commission about the state’s compliance with a directive. Nevertheless, the small number of such cases supports the argument that pre-existing domestic legislation is rarely sufficient to fulfill all EU requirements contained in an EU directive. Instead, if pre-existing measures are followed by domestic instruments adopted only after the transposition deadline, the odds that the Commission will open an infringement case are 83% higher than if timely measures (enforced after the directive adoption but before the transposition deadline) follow pre-existing ones. Such cases indicate situations where a member state did not manage to modify its legislation on time.

Another reason why pre-existing measures are associated with non-compliance is because they may require extensive adjustments in order to fully meet the EU requirements or they may simply be of only partial relevance to the transposition process. As already discussed in the measurement section, the extent to which pre-existing measures actually fit the EU directives is captured by the extent of adjustments to national legislation that is needed to transpose the directive at hand. In model 3, we report the results from the separate categories of early measures policy fit on the likelihood of reasoned opinion relative to not sending such measures. As expected, notifying low fitting early measures is associated with most non-compliance problems (increasing the odds of non-compliance by over 400%), whereas the effect of early measures considered to be highly fitting is not significantly different from starting the transposition process following a directive’s adoption. This finding is in congruence with H2b that early measures are only partially relevant and lead to compliance problems if they require extensive adjustments.

Finally, model 4 in Table 1 tests the prediction that transparency requirements specified in a directive’s text decrease differences between low and high fitting early measures in their impact on the likelihood of an infringement case (H3). This is because transparency requirements signal that the other member states put high emphasis on the proper transposition of a directive, which is expected to induce governments to pay the costs for making their legislation compliant with the respective directive. To test this hypothesis, in model 4 we apply the analysis to the sample of cases where a member state reported at least one pre-existing measure. First, we aggregated the four different categories of early measure policy fit into one ordinal variable with higher values indicating fewer adjustment costs being necessary for compliance. We then interacted this measure with the transparency requirement variable.

Based on the analysis, a member state that sends a low-fitting early measure is less likely to get a reasoned opinion for directives that require member states to submit a concordance table as part of the notification procedure. This is shown by the negative coefficient of transparency requirements in model 4, which indicates the effect of this variable when policy fit is zero (or extremely small). In turn, model 4 shows that the effect of policy fit on infringement proceedings is also negative and significant for directives that did not specify a transparency requirement (Transparency requirement = 0). Put differently, for directives without transparency requirements the likelihood that such cases will get an infringement proceeding for non-compliance decreases as the policy fit of early measures increases. However, the positive interaction effect shows that the negative relationship between policy fit and reasoned opinion is significantly weakened for directives with transparency requirements. To illustrate this better, we plotted the predicted probabilities of an infringement case (a reasoned opinion) given different levels of early-measure policy fit for directives with and without transparency requirements. Figure 2 shows that for directives with transparency requirements, member states have a relatively low probability of getting a reasoned opinion, irrespective of the level of policy fit of their reported pre-existing measures. OPEN IN VIEWER

To summarize, we find no support for H1 but evidence for H2a, H2b and H3. These findings are stable across different model specifications, such as excluding control variables, and are robust to including policy area and member state fixed effects. In particular, reporting pre-existing national legislation to indicate compliance does not portray a country’s actual leadership in EU integration but is in fact most often related to compliance problems. In addition, as expected, the results show that the extent to which reporting pre-existing measures ends up in detected non-compliance problems of a member state depends on: (1) whether the member state manages to adjust its pre-existing legislation on time; (2) how fitting the state’s pre-existing legislation was relative to the directive; and (3) how important the transposition of a given directive is for other actors (as reflected in the inclusion of transparency requirement in the directive’s text). The latter could be because increased transparency dissuades member states from just notifying partially relevant policies.

One may view the results with caution given our reliance on the Commission’s infringement cases to capture member states’ non-compliance. As we discussed earlier, infringement proceedings are affected by the capacity of the Commission to detect non-compliance problems and its willingness to enforce their resolution. Thus, the lack of an infringement case may indicate either that the member state did in fact comply or that the Commission failed to detect cheap signals that were not compliant. This is not a big problem in the present study because if the Commission is not able to detect some of the pre-existing measures as related to non-compliance problems, this should blur the positive relationship between sending such measures and infringement cases, making the link even non-significant. In a sense, the infringement procedure thus provides a conservative test of the link between sending pre-existing measures and non-compliance, making it more difficult for us to find a positive relationship.

Our findings are robust to the inclusion in the models of a number of control variables, which could have affected the relationship between reporting pre-existing measures and receiving a reasoned opinion. Looking at the control variables, we find no evidence that more influence during the decision-making stage, higher voting weights in the Council of Minsters or holding the rotating Council Presidency at the time of directive adoption, decreases the likelihood that the Commission would open an infringement case against a member state. Governmental EU support does not influence the likelihood of infringement cases, while governmental effectiveness has a significant negative effect on a member state’s likelihood of receiving a reasoned opinion only in model 3. As expected, the likelihood of compliance problems increases if there was a change in government before the completion of the transposition process. In congruence with recent findings, old member states are significantly more likely to receive a reasoned opinion from the Commission than the new member states (Knill and Tosun, 2009; Sedelmeier, 2008; Yordanova and Zhelyazkova, 2011), although this effect is not robust in all models. Finally, the more recitals directives contain, reflecting the directives’ level of complexity, the more likely it is that the Commission will open infringement cases against non-compliance.

Conclusion and discussion

In this article, we have explored the relationship between the notification of implementation measures by the European Union member states and detected non-compliance by the European Commission. We have argued that member states have incentives to engage in ‘cheap talk’ that could be associated with compliance problems with EU law. One strategy to shirk from EU obligations is to report pre-existing national laws that may be only partially, if at all, relevant for the transposition of a particular directive. Whereas we do not expect that the European Commission will tend to blindly accept these measures as meeting the transposition requirements but instead that it usually acts by issuing an infringement case, reporting old measures may still serve as a time winning strategy for short-sighted governments until costly sanctions are applied. However, sending non-compliant pre-existing measures would be costly and incur high reputational costs on a national government if directive transposition is of high importance for the other member states and/or the Commission. In such cases, we expected member states to have strong incentives to fully comply with the directive’s requirements.

The analysis of this article showed that, indeed, contrary to general expectations, notifying pre-existing measures is most often related to compliance problems and triggers the opening of infringement cases by the Commission. As we hypothesized, this relationship is weaker the more the pre-existing national legislation actually fits the requirements of an EU directive or when the EU directives stipulate requirements for transparency in the national implementation process.

Our findings have important implications for research on member states’ EU compliance. First, whereas it is generally agreed in the compliance literature that reporting measures enforced before the adoption of an EU directive signifies that a member state is a driver of European integration or that its legal system closely matches the directive’s goals, our results question these assumptions by showing that pre-existing measures are more likely to be related to compliance problems than reflect member states’ actual transposition effectiveness. Second, this article contributes to discussions about the proper measurement of transposition success. In particular, the results of this study lend support to critiques against the use of the very first notified legal instrument to measure actual non-compliance with EU directives (König and Luetgert, 2009). We showed that this is problematic not just because early measures signify only the start of the transposition process, but also because they may be related to member states’ incentives to shirk from their EU obligations. Finally, and more substantively, our findings suggest that the implementation process improves whenever the European Commission or the member states consider compliance with a directive important and clearly signal that in order to generate an expectation of high reputational cost for not (fully) complying with a directive (e.g. with a concordance table requirement).

We conclude with a few suggestions for future research. First, alongside pre-existing legislation, timely, or even late, measures may sometimes constitute ‘cheap talk’ too because reporting such measures does not tell us anything about the extent to which they meet the EU requirements. This renders necessary further research to establish when such self-reported measures are likely to be related to compliance problems. Second, it would be useful to study to what extent member states’ incentives to report untruthful information are driven by the ability of the Commission to detect non-compliance problems and its willingness to pursue infringements. However, only an exogenous measure of non-compliance would allow examining when ‘cheap talk’ associated with implementation problems effectively avoids infringement cases. While still in its infancy, research aimed at uncovering such new compliance indicators is promising (see Falkner et al., 2007; König and Mäder, 2013; Zhelyazkova, 2013). Finally, further research is needed to establish when and why member state governments resort to ‘cheap talk’ without actually complying with their international obligations.