Unpacking the Unilateral Actions in Latin America Presidentialism

• In Event: Political Elites and Policymaking

Abstract

This paper analyzes from a comparative perspective the use of unilateral actions by Latin American presidents executed through administrative decrees. We use an unsupervised machine learning technique (structural topic models - STM) to recover the topics of presidential decrees and demonstrate that presidential attention lies in a great variety of policy areas with different potential goals driving their behavior. Based on the textual analysis done with the STM we recover the topics covered in all decrees and estimate a topic prevalence for each document and differentiate decrees by their goals, whether they create, expand, reduce, or extinguish previous programs or agencies in different policy areas. We built a novel and large dataset that comprises administrative decrees issued by presidents in seven countries (1985-2018): Argentina, Brazil, Chile, Paraguay, Colombia, and Peru. Making use of document metadata (document-level covariates), such as presidential mandate; ideological distance president-median legislator; presidential powers, we can better estimate the topical prevalence and improve the overall quality of estimated topics.

Authors

• Magna Inácio, Federal University of Minas Gerais

• Filipe Recch, Stanford University