When and why high civil servants demand information from interest groups in policymaking. A Southern European perspective

1. We know that the focus on high-level civil servants could be disputed because some of them are managers or simply higher-level advisors of ministries. However, there is an established stream of research that has provided empirical evidence that the main role of these civil servants is supporting politicians in designing, monitoring and assessing policies as well as implementing specific policy programmes according to the organisational characteristics of the related state (Hammerschmid et al. Citation2016; Kuperus & Rode Citation2016; Steen & Weske Citation2016).

2. Albareda and colleagues (Albareda, Braun & Fraussen Citation2023) take into account policy capacities to explain interest groups’ influence, but they refer to interest groups’ policy capacities rather than civil servants’ policy capacities.

3. On this subject, we can add a pragmatic consideration: since it makes little sense to ask civil servants for their perception of the whole national administrative system’s analytical capacities, we did not insert this kind of question into the survey on which this study is based. This would have meant that treating systemic analytical capacities as an independent variable would also mean using incoherent empirical indicators (organisational and individual capacities on the basis of the respondents’ perceptions, whereas systemic capacities were considered on the basis of external sources).

4. In this way, we can also treat ‘interest groups’ and ‘stakeholders’ as synonymous. However, to check the robustness of our results, in the Supplementary Materials, we also test multivariate regressions with representative associations only: see, in particular, Table SM13.

5. We opted for an additive index because single responses are highly correlated (Pearson correlations between 0.28 and 0.49, always at the highest level of statistical significance). However, to further confirm the methodological correctness of our choice, we also ran a principal component analysis among the various items. See, for more details, both Table SM5 (bivariate correlations) and Table SM6 (principal component analysis) in the Supplementary Materials.

6. We cannot exclude the possibility that the wording of this question may (also) refer to considerations at the systemic rather than the organisational level. However, we are fairly confident that in the vast majority of cases, high-level civil servants who responded did so on the basis of their perceptions of the specific administration in which they worked rather than the overall national administrative system.

7. Single responses are highly correlated (Pearson correlations between 0.25 and 0.67, always at the highest level of statistical significance) and are all associated with an underlying factor/component that explains more than 51% of the total variance. See, for more details, both Table SM7 (bivariate correlations) and Table SM8 (principal component analysis) in the Supplementary Materials.

8. Single responses are highly correlated (Pearson correlations between 0.08 and 0.73 and statistical significance at the 0.027 and 0.000 level, respectively). Furthermore, all specific items but one (that about the familiarity with online consultations) are associated with an underlying component that explains approximately 43% of total variance. Thus, we use the simple additive index in the main text, whereas we run a robustness test with a different independent variable, built on the basis of the results of the principal component analysis among various items, in the Supplementary Materials. See, for more details, Table SM9 (bivariate correlations), Table SM10 (principal component analysis) and Table SM16 (robustness test) in the Supplementary Materials.

9. For detailed country data on systemic analytical capacities, see Supplementary Materials, Table SM11. However, there are many potential indicators of systemic analytical capacities. Thus, in the Supplementary Materials, we checked our results also recurring to the World Bank indicators of governance effectiveness; the empirical findings did not change (see Table SM14). Furthermore, to verify whether the empirical results are driven largely by the Italian case, by far the most represented in our sample, we run a further robustness check (see Table SM17) with country dummies; again, our findings are confirmed.

10. Since our dependent variable is an additive index ranging from 5 to 25, we are quite confident that it can be treated as a continuous variable: this is why we opted for OLS as the preferred statistical method. However, to check our results, in the Supplementary Materials, we also run an ordinal regression on the dependent variable. The empirical results do not vary much (see, in particular, Table SM12).

11. In the Supplementary Materials, we recur to an ordinal regression instead of OLS regressions (Table SM12), we use World Bank indicators instead of the Bertelmann-Stiftung index (Table SM14), and we utilise variables that originate from principal component analyses run on all items taken into account (Table SM16).

12. Regarding the details of how we grouped ministries in the three categories, see Table SM3 in the Supplementary Materials.