Abstract
This paper examines variation in the timing of compliance with European directives. It formulates and tests the hypothesis that member states' policy-based incentives to deviate from the content of directives influence delay in compliance. This hypothesis is tested along with other factors that are posited to influence compliance, including the amount of discretion directives give member states, the level of misfit between national and European-level laws, and characteristics of member states. The hypotheses are examined in a quantitative research design using arguably the best available information on compliance: national responses to six labour market directives investigated by Falkner et al. (Citation2005) for Complying with Europe. The present study develops Falkner et al.'s analysis in two respects. First, it identifies new theoretically important variables and offers measures of these, notably member states' policy-based incentives to deviate and the amount of discretion granted by directives. Second, it tests these hypotheses using multivariate analysis, while Falkner et al. applied bivariate tests only. In contrast to Falkner et al.'s conclusions, the findings indicate that misfit between national and European laws significantly reduces the likelihood of timely compliance. While political opposition at the time of a directive's adoption is not linked directly to compliance, member states tend to oppose directives that do not fit existing national laws. Compliance is more timely for directives that grant more discretion.
Acknowledgements
I am grateful to Gerda Falkner, Miriam Hartlapp, Simone Leiber and Oliver Treib for sharing their data on the timeliness of compliance. I thank the Institute for International Integration Studies at Trinity College Dublin for research support, and Matthew Wall for research assistance. Two anonymous reviewers and René Torenvlied provided constructive comments on an earlier version.
Notes
1. reports Cox regressions using the exact-partial calculation for dealing with ties (Cleves et al. Citation2004: 142). This calculation assumes that the tied observations, cases with the same duration before compliance, really did occur after the same duration following the deadline. Two other ways of dealing with tied observations, the marginal calculation and the Breslow approximation, produced almost identical results.
One of the most important assumptions of Cox regression is the proportional hazards assumption. This assumption is not violated by these data. The assumption is that the effect of a change in each independent variable is proportional and constant over time (Box-Steffensmeier and Jones Citation2004: 131–2). The assumption was tested in three standard ways. First, a link test was performed on each of the two models presented in (Cleves et al. Citation2004: 175). Second, the Grambsch and Therneau global test was applied to both of the models. This test indicates whether there is evidence of non-proportional hazards in the model as a whole (Box-Steffensmeier and Jones Citation2004: 135). Third, Harrell's rho tests were performed on each independent variable separately. Not one of these tests gave a significant p-value, indicating that there is no reason to reject the proportional hazards assumption.
2. A separate Cox regression (not reported) was performed including the variable ‘states’ disagreement with directives' and excluding the ‘misfit’ variables, and the variables relating to discretion and the length of the directives. ‘States' incentive to deviate’ was still not associated with a significant effect.