Classifying Local Government Commitments to Equity: A Machine Learning Approach

• In Event: Equity and Access

Abstract

Disparities in service-delivery, economic opportunities and living conditions are perhaps the most pernicious and destabilizing cluster of problems confronting urban regions today. While cities pursuing urban sustainability policy agendas have drawn increased attention to issues of social equity, research suggests that achieving goals of equitable outcomes, especially in low-income and communities of color, ranks below other sustainability-themed priorities such as economic development, fiscal efficiencies and environmental conservation.

The call to determine how cities approach social equity has grown louder as police violence and the disproportionate toll of the COVID-19 pandemic on communities of color have highlighted long standing inequities at the local level. These inequities are compounded by mounting evidence of environmental injustices and the potentially devastating effects of climate change on the fiscal, social, and environmental conditions of people of color and low income communities in urban areas. Even with undeniable urgency, determining how cities address social equity in sustainability efforts is complicated by the multidimensional conceptualization of sustainability and outcome definitions and measurements.

Synthesizing policy and resource-based theories of organizational capabilities, this paper adapts an ensemble machine learning (ML) strategy to analyze U.S. city policy- and capacity-commitments to social equity actions. Social equity can be defined and operationalized in many ways, such as through the distribution of benefits or public goods, the procedures for inclusive policy making, or the acknowledgement and recognition of past injustices which influence outcomes today. Both policy choices and resource-commitments hinge on these foundational, problem-oriented distinctions. From a public administration perspective, such conceptualizations share a common dependency on organizational resources, but differ significantly in how they should be deployed. This study examines how policy commitments to equity influence such local government resource-allocation decisions. Using sustainability and economic development surveys, social vulnerability indicators and other administrative data, we classify cities based on their commitments to equity in several interrelated planning, capacity-building and performance actions or activities. Utilizing random forest, gradient boosting, and neural network predictive models for classifying specific outcomes, we then determine which commonly utilized environmental, socio-economic or institutional variables are most important for predicting a range of specific equity-targeted programmatic actions and activities. By doing so, we examine the potential that nonlinear and interactive community and organizational features can better predict the resource commitments local governments must make to improve equity outcomes. ML applications have gained recent popularity as computational advancements and the proliferation of data make it possible to glean novel insights from complex phenomena. Using an ML-based estimation strategy, the study provides important insights into the role of both policy commitments and administrative behavior in equity outcomes.

Authors

• Aaron Deslatte, Indiana University Bloomington

• Cali Curley

• Eric Stokan, University of Maryland, Baltimore County

• Kathryn Elizabeth Wassel