Opinion Change or Differential Turnout: Changing Opinions on the Austin Police Department in a Budget Feedback Process

TL;DR

This study treats the George Floyd–related protests as an exogenous shock to Austin's online budget-feedback platform, revealing a dramatic shift toward police funding reductions that persists into 2021 and widens the opinion gap. It leverages a knapsack-style budgeting task, demographic analyses, and three-cluster analyses to identify persistent opinion profiles and to compare with Participatory Budgeting elections. The follow-up survey and cross-year clustering indicate that the observed shift reflects real opinion change rather than turnout alone, and that clustering can illuminate opinion minorities and improve equity in civic feedback. The findings inform the design and evaluation of online budgeting tools and demonstrate how major social events can reshape public opinion and participation patterns.

Abstract

In 2020 the tragic murder of George Floyd at the hands of law enforcement ignited and intensified nationwide protests, demanding changes in police funding and allocation. This happened during a budgeting feedback exercise where residents of Austin, Texas were invited to share opinions on the budgets of various city service areas, including the Police Department, on an online platform designed by our team. Daily responses increased by a hundredfold and responses registered after the "exogenous shock" overwhelmingly advocated for reducing police funding. This opinion shift far exceeded what we observed in 14 other Participatory Budgeting elections on our Participatory Budgeting Platform, and can't be explained by shifts in the respondent demographics. Analysis of the results from an Austin budgetary feedback exercise in 2021 and a follow-up survey indicates that the opinion shift from 2020 persisted, with the opinion gap on police funding widening. We conclude that there was an actual change of opinion regarding police funding. This study not only sheds light on the enduring impact of the 2020 events and protests on public opinion, but also showcases the value of analysis of clustered opinions as a tool in the evaluation toolkit of survey organizers.

Figures (38)

• Figure 1: Responses per day (2020) on a logarithmic scale
• Figure 2: Design of revenue portion of the 2020 exercise.
• Figure 3: Design of expenditure portion of the 2020 exercise.
• Figure 4: Responses per fee category, what level of fee increases (revenue) would be supported. Background color indicates size of support. Significant means 3% or more.
• Figure 5: Responses per service area, what level of expenditure increases would be supported. Background color indicates size of support. Significant means 3% or more.