Coordination and Decentralization in Provision of Social Services

Poverty is a prevalent issue in the United States, impacting over twenty percent of the population. Given poverty's detrimental impact on communities, there is a pressing need to predict hardship as well as provide timely and effective interventions. Despite the multi-dimensional nature of disadvantage -- ranging from housing instability and lack of access to physical and mental health care to educational support and food security -- agencies providing services often specialize in a single area, and seldom coordinate with other agencies. This decentralized nature of assistance, policy-makers and advocates argue, leads to individuals’ needs not being addressed holistically, and may perpetuate cycles of poverty. 

In this talk, we focus on two key questions: in the first part, we study the problem of evaluating such decentralized systems of provision of social services. Under natural assumptions on peoples’ utilities for services as well as fairness considerations, we find that decentralized systems of allocation, even when optimal, can have a large gap when compared to optimal global provisions. We then consider coordination mechanisms, inspired by insights from work with California’s Sonoma County system, and theoretically examine whether these interventions can reduce gaps between local and global systems of provision. In the second part of the talk, we discuss recent work on predicting poverty. We evaluate popular conjectures around what makes poverty challenging to predict, and provide evidence that a key issue may be how we measure poverty and set guidelines for determining eligibility for receiving services. We will then discuss research directions where algorithmic and computational techniques can contribute to policy and advocacy efforts.