Preference alignment in multiparty governments: control of public policy in Brazilian presidentialism

Abstract

The formation of multiparty governments increases the risk of asymmetry of information between principals and agents so that the cabinet's policy conflicts are reflected in the behavior of parties in parliament. Several studies reveal that mutual control between cabinet parties is a way to compensate for the loss of information inherent in the delegation. While analysts often focus on the policy-making process, I argue that there is a more diverse set of strategies that allow parties to scrutinize policies implemented by their cabinet partners. To test this argument, I analyze the multiparty governments formed in Brazil between 1995 and 2014. I use network analysis and machine learning—Gibbs Sampling Method, Bayesian algorithm derived from Latent Dirichlet allocation (LDA)—to show that, in situations in which ministerial portfolios are distributed to actors with different policy preferences, parties intensify the use of Information Requirements (RIC), an important tool for horizontal control, to monitor the ministries, policies, and programs that interest them. The structure of the intra-cabinet control networks varies according to the importance of the ministries: the parties allocated in the ministries with the largest portion of the federal budget—those with the most resources for the implementation of public policies—are the actors with the greatest degree of centrality in the intra-cabinet mutual control networks.

presidentialism; government coalitions; public policy; network analysis; machine learning