(De-)Centralisation and Attribution of Blame and Credit

Abstract

Assigning responsibility for political outcomes is complex in political systems where power is apportioned across local, regional and national levels of government. Does attribution of responsibility reflect the allocation of formal authority, as implied by literature on delegation and blame avoidance, or is it always the national level government that gets the blame when the media focus on failures and unpopular policy outcomes? In this article, I develop a set of hypotheses about public attribution of responsibility within a policy sector characterised by multiple tiers of government involvement. The hypotheses are tested using a novel dataset based on a content coding of responsibility attribution in more than 2000 newspaper articles about health care issues in Denmark. Decentralisation of formal authority does show a blame deflection effect in media coverage of health care issues, but the analysis also points to several factors that moderate the effect. In that way, the article advocates a renewed interest in the often invoked relationship between the institutional allocation of authority and public attribution of responsibility.

Key Words:

• Decentralisation
• blame avoidance

Acknowledgements

This research has received financial support from the national Danish Council for Independent Research, Social Sciences (FSE). The author gratefully acknowledges the comments of Jens Blom-Hansen, Christoffer Green-Pedersen, Søren Serritzlew, and the anonymous reviewers on earlier versions of this article. The author would also like to thank Ann-Sofie Tranæs and Morten Barsballe Schmidt for excellent research assistance. The usual caveat applies.

Notes

1. On negativity bias, see for example, Lau (1985) and Soroka (2006).

2. This is probably also why media coverage has been a popular data source in many previous blame avoidance studies (see for example, Brändström et al. 2008, Hood et al. 2009, Sulitzeanu-Kenan 2007).

3. The use of the term ‘delegation’ is justified by the fact that the present discussion is limited to the range of political systems where simple majorities at the national level can abolish the county/regional policymaking level without facing constitutional constraints, which implies that the policymaking power of these subnational institutions is delegated from the centre and hence corresponds with, for instance, Flinder’s (2008) definition of delegated governance.

4. The minister of health is replaced at the end of the period, but the statistical analysis controls for this change.

5. Danish newspapers are not overly partisan, but historically they have different political ties.

6. A large number of articles were excluded because they turned out to be about something other than Danish health care. Furthermore, in line with traditional measures of a ‘media agenda’ (see Dearing and Rogers 1996, p. 35) we focus on news stories and exclude more subjective items such as discussion articles, editorials, feature articles, letters to the editor and reviews. Finally, a number of duplicate articles published in different but cooperating newspapers were excluded.

7. To take into account the hierarchical structure of the data (articles within newspapers) I use robust clustered standard errors (RCSE) (see Williams 2000). Another methodological question is the use of fixed effects (FE) versus random effects (RE) models. If there are unobserved year effects, the within (or FE) estimator should be used, but using only within variation leads to less efficient estimation. However, based on the conventional Hausman test for fixed effects it cannot be rejected that the RE estimator is fully efficient and hence I continue with a random effect estimation in the models reported in Table 4.

8. Two alternative estimation procedures have been shown to produce unbiased standard errors in analyses of clustered data: robust clustered standard errors (RCSE) (see Williams 2000) and multilevel models (see for example, Goldstein 2003). Given the controversy of this issue the analyses in Table 4 have been re-estimated using a multilevel estimation. This re-estimation, however, leads to essentially the same conclusions as those obtained by using the RCSE approach (results available from the author upon request).

9. In addition, I controlled for the potential ‘responsibility release’ from a replacement of the minister of health during the period of investigation. The new minister of health took office on 27 November 2007. The previous minister of interior and health was promoted to minister of finance. The effects of the replacement are statistically insignificant, however, and do not change the conclusions derived from Table 4 (results are available from the author upon request).