Abstract
This study focuses on a central question in the literature on policy agendas and punctuated equilibrium: why are some agendas more punctuated than others, and what causes these punctuations? In particular, is it friction – wherein barriers to change lead to the build-up of tension that finally overflows – or rather cascades that occur owing to positive feedback loops as actors imitate other actors? We hypothesize that both are at work, and that under certain conditions – e.g., the number of actors and the amount of communication between them – one mechanism is stronger than the other. We test our hypotheses with data on parliamentary activities (interpellations and oral questions) and media coverage in Belgium in the 1990s. We find evidence of both friction and cascading contributing independently to the typical punctuated pattern of policy agendas.
ACKNOWLEDGEMENTS
We thank Brandon Zicha and Peter Mortensen, as well as the JEPP referees, for useful comments on earlier versions of the paper.
Notes
Belgian newspapers are not issued on Sundays. Therefore, for the various television channels a mean score for each issue for Saturdays and Sundays is calculated to substitute the original Saturday score, while the Sunday score is deleted in order to keep data comparable with the newspaper data. Furthermore, newspaper data for Tuesdays and Thursdays, that are typically not coded, are estimated based on previous and subsequent values. Results in terms of dependencies and imitation across outlets is not affected by this imputation.
One could argue that it is not necessary to include issue-specific intercepts, since differencing the original attention score variable already removes large part of the variation across issues. However, given the fact that we have a large dataset at our disposal, we consider it a ‘safe’ choice to conduct fixed-effect analyses. Furthermore, our data might not meet all the assumptions for linear regression. Especially the normal distribution of errors might be problematic – after all, we already anticipate them to be leptokurtic. However, regression analysis is not very sensitive to violations of this assumption (Hayes Citation2005: 298). Additionally, problems with non-normality relate to the estimation of the accuracy of confidence intervals of parameters. As we will argue, our substantial interest is in the distribution of the errors, rather than in the causal relationships that are estimated.