A reputational perspective on structural reforms: how media reputations are related to the structural reform likelihood of public agencies

Abstract Despite recurrent observations that media reputations of agencies matter to understand their reform experiences, no studies have theorized and tested the role of sentiment. This study uses novel and advanced BERT language models to detect attributions of responsibility for positive/negative outcomes in media coverage towards 14 Flemish (Belgian) agencies between 2000 and 2015 through supervised machine learning, and connects these data to the Belgian State Administration Database on the structural reforms these agencies experienced. Our results reflect an inverted U-shaped relationship: more negative reputations increase the reform likelihood of agencies, yet up to a certain point at which the reform likelihood drops again. Variations in positive and neutral reputational signals do not impact the reform likelihood of agencies. Our study contributes to understanding the role of reputation as an antecedent of structural reforms. Complementing and enriching existing perspectives, the paper shows how the sentiment in reputational signals accumulates and informs political–administrative decision-makers to engage in structural reforms.