Research has suggested numerous other variables that influence transposition delay and speed. These factors must be controlled for in the analysis presented below. The majority of the control variables in this study fall into one of three categories, “features of the directive and Commission monitoring,” “national variables,” and “transposition process-related variables” (see Steunenberg and Rhinard 2010, 500-501).
Five variables fit within the “features of the directive and Commission monitoring” category. The first variable in this category is the length of time member states are granted to transpose a directive. If difficulties in transposing directives appear likely, member states may be granted a longer period of time to transpose a directive. Thus, one would expect directives to be transposed more slowly when member states are given a longer period of time to transpose (see Kaeding 2006; Steunenberg and Rhinard 2010). This variable is measured as the number of weeks granted between the adoption of a directive by the EU, and the directive’s transposition deadline.
A second variable within this category is the complexity of a directive. When a directive is particularly complex member states may have difficulty understanding exactly what measures are necessary to transpose it (Mastenbroek 2003; Kaeding 2006). Therefore, a greater degree of complexity should result in slower transposition. Complexity is operationalized as the number of recitals in the preface of a directive (see Kaeding 2006).
A directive’s status as either new or as an amendment to a previous directive has also been found to influence delay in and the speed of transposition. Directives that amend previous directives are simply adaptations of existing policies, and should receive less opposition from member states (Mastenbroek 2003; Borghetto et al. 2006). This leads to an expectation that amendments to previous directives will be transposed more quickly. This variable is operationalized as a dummy variable, with a 1 denoting whether the directive is new, and a 0 if it is an amendment to a previous directive.
A fourth variable in this category concerns whether a directive was issued solely by the European Commission or if the European Council and/or the European Parliament were involved. Those directives that are issued solely by the European Commission through the delegation of policymaking power tend to elaborate on existing directives, making them easier for member state governments to transpose. Therefore, directives adopted solely by the European Commission should be transposed more quickly (Mastenbroek 2003; Steunenberg and Rhinard 2010). This variable is operationalized as a dummy variable with a 1 denoting a directive that was adopted solely by the European Commission, and a 0 denoting a variable that was adopted through the legislative process.
The final variable in this category is the intensity of European Commission monitoring. Research has suggested that the degree to which the European Commission monitors and notifies member states of infringements in the transposition process can affect the speed of transposition, with greater monitoring leading to faster transposition (Börzel 2001; Tallberg 2002; Steunenberg and Rhinard 2010). This variable is measured as the “average number of formal letters of notification sent by the Commission to member states in a specific year, normalized to the [0,1]-interval” (Steunenberg and Rhinard 2010, 502). The intensity of Commission monitoring, however, changes over time, making this variable is a time-varying covariate.
Within the second category of “national variables,” many previous studies have included public opinion. As this is the main independent variable in this study, a measure of public opinion is already included in all models. Recent research has also included country dummy variables in the analysis (see Steunenberg and Rhinard 2010). Therefore, dummy variables for the United Kingdom, Germany, Spain, and Greece are included in this analysis with the Netherlands as the baseline category.
The first variable in the final category, “transposition process-related variables,” concerns the number of veto players in a country. The more veto players involved in the transposition process, the
Christopher J. Williams
slower transposition should become (Kaeding 2006; Steunenberg and Rhinard 2010). This variable is operationalized in a similar manner to Steunenberg and Rhinard (2010), who created an index that varies by the transposition procedure being used by a member state (i.e. legislative, ministerial, or cabinet level). Originally, Steunenberg and Rhinard coded this variable as an additive variable. If the procedure being used to transpose a directive was at the ministerial level, the number of government ministries involved in transposition was used as a measure of veto players. If the procedure was at the cabinet level, an index indicating the autonomy of the prime minister (derived from scores originally developed by Bergman et al. 2003) was added to the number of government ministries involved in transposition. If the procedure being used was legislative, the measure included the number of government ministries involved in transposition, plus the prime minister’s autonomy index, plus the member state’s score from Tsebelis’ (2002) veto player index.
A problem arises with this coding, however. Tsebelis’ veto player index is missing data concerning Greece. This affects the coding of Steunenberg and Rhinard’s additive veto player index for directives that are transposed under a legislative procedure. To address this issue, a new additive veto player index is created. It is nearly identical to that used by Steunenberg and Rhinard, however, Tsebelis’ measure is replaced with the number of parties in government at a particular time. This new index does not appear to differ substantially from Steunenberg and Rhinard’s original additive index in the Netherlands, Germany, the United Kingdom, and Spain.7 This index’s value changes over time,
making this measure a time-varying covariate.
A second variable in this category is member state experience with transposition. Previous literature has shown that experience with transposition is an important factor influencing transposition speed and timeliness. As member states become more experienced with transposition, they become better at it, and it occurs more quickly (Tallberg 2002; Steunenberg and Kaeding 2009). This variable is operationalized as the length of a state’s membership in the EU. This measure changes with time, thus, this variable is also a time-varying covariate.
As a final variable in this category, national elections are controlled for. It has been theorized that national elections can slow transposition as those involved in the transposition process hope to avoid “criticism during political campaigning” (Steunenberg and Rhinard 2010, 504). This variable is operationalized as a dummy variable with an observation being coded as a 1 if a national election is occurring during a Eurobarometer semester. This variable changes by Eurobarometer semester, thus, this is also a time-varying covariate.
Beyond the variables in the three categories mentioned above, a number of other control variables are also included. Differences exist in transposition delay across policy areas (Berglund et al. 2006). Therefore, dummy variables for the different directive policy sectors (food legislation, social policy, transport, and utilities regulation) are included in this study. Social policy is used as the baseline policy domain. A variable denoting the exact length of each Eurobarometer semester is also included. The Eurobarometer semester is roughly a half-year, however, the exact length of each semester can vary, thus, it is imperative to control for the exact length of each semester in months.
4.3. Data
The data used in this study concerning the time until transposition is the same data used by Steunenberg and Rhinard (2010). This data concerns directives adopted by the EU in the policy areas of food legislation, social policy, transport, and utility regulation between 1 January 1978 and 1 January 2003. This end date is used in order to avoid cases in which member states may still be attempting to transpose directives (see Steunenberg and Rhinard 2010).
These policy areas were chosen as they, “differ sufficiently in terms of the time at which they were developed at the European level, allowing for differences in policy characteristics and national transposition experiences” (Steunenberg and Rhinard 2010, 505).
The data further focuses on transposition by the governments of the Netherlands, Germany, the United Kingdom, Spain, and Greece. This set of countries:
7 The Pearson’s R correlation between the new additive veto player index (using the number of parties in government), and
Steunenberg and Rhinard’s original additive veto player index is 0.97.
Responding to Euroscepticism
...includes some of the founding member states of the Union as well as more recent members (UK in 1973, Greece in 1981, and Spain in 1986). Moreover, these countries display substantial variation on transposition performance, as indicated by the Commission’s transposition ‘scoreboards’ over the last couple of years, covering most of the variation that can be found between EU member states (Steunenberg and Rhinard 2010, 505).
In total, there are 1,160 directive-state dyads in this dataset, representing 317 distinct directives. Of the full sample, 251 of the dyads concern the Netherlands, 234 of the dyads concern Germany, 234 of the dyads concern the United Kingdom, 221 of the dyads concern Spain, and the final 220 dyads concern Greece.
The structure of the original data used by Steunenberg and Rhinard was transformed to account for the time-varying nature of euroscepticism. The level of aggregate euroscepticism in each country changes with each new Eurobarometer survey. Therefore, each observation in the original dataset is expanded by the number of Eurobarometer semesters that passed between the publication of a directive by the EU, and the subsequent transposition date of that directive in each member state.8
This results in a total of 5,518 observations.