Mobilizing Media: Estimating the Impact of Participatory Budgeting on Community-Level Political Expression

This paper, part of a broader multi-method dissertation project exploring community level impacts from localized processes, presents the method and results of a new approach to describing macro impacts from a local participatory budgeting (PB) process in the United States. Deliberative democratic innovations like PB have often been motivated by the hope that these local institutional changes can serve as a catalyst for a broader transformation of the political culture in a community – the attitudes that people have toward government and their self-perceptions as people and a community that can work together for positive or creative change. However, defining and measuring changes in political culture is an ongoing challenge for social scientists, especially when comparing communities across time and place. Common research methods like surveys or extended fieldwork in the community can be hard to implement in multiple places at once, and neither can take us back in time to get good historical trends. In this paper I present a new approach that allows partial measurement of changes in political culture, using political discussion in local newspapers as an approximation of the common beliefs and expectations the public hold with respect to the government and others in their community. I implement a new supervised machine learning approach to capture changing expressions of political culture across time and in multiple communities. I create a dataset of over 20,000 local newspaper articles from 2004-2015, including political reporting, opinion and letters to the editor in two cities struggling with bankruptcy and local investment in California: Vallejo, which implemented an ongoing PB process in 2012, and Stockton, which did not implement a PB process upon exiting bankruptcy. With a team of undergraduate research assistants, I classify a sample of articles into qualitative categories of mobilizing or demobilizing political expression (with additional categories for neutral or irrelevant reporting). I then use the ReadMe package in R to generate estimates of how the tenor of political speech changes over time in the whole set of articles in both cities. This methodology provides an opportunity to explore whether PB can be seen to have provoked a cultural shift toward a more proactive, mobilizing trend in public political expression. The quantitative results are, in turn, contextualized within a qualitative narrative that is the product of interviews and observation over about four weeks of PB activities in Vallejo in 2013 and 2014. This paper thus makes two primary contributions of interest to the study of macro-political impacts of deliberation: On the one hand, it provides a focused empirical analysis of the impact of a specific form of deliberative participation, adding another 'data point' to the growing field of scholarship on macro-level impacts. On the other hand, it introduces and explains a method of text analysis that provides new opportunities for comparative and/or historical analyses of community impacts. The method described here could be readily extended into other contexts or specific discourse categories.