M.S. Applied Data Science - Capstone Chronicles 2025

3

3 Literature Review (related works) 3.1 Disability and Labor-Market Barriers She and Carter (2023) show that employment limitations are the largest contributor to poverty among disabled adults. Their work highlights that workplace exclusion and inaccessible job structures persist even for those with advanced degrees. They argue for integrated datasets linking disability, labor force participation, and education — precisely the approach undertaken in this project. 3.2 Educational Attainment and Socioeconomic Disadvantage Kim, Torres, and Zhao (2022) find that while higher education typically lowers poverty risk, the effect is significantly weaker for marginalized populations, including the disabled. Their regression analysis across U.S. counties underscores the need to account for interaction effects between social identity and structural opportunity, a framework mirrored in this capstone’s modeling design. 3.3 Spatial Inequality and Economic Mobility Chen and Rhee (2020) examine ZIP- and tract-level disparities in economic mobility and identify strong geographic clustering of poverty. Their work supports this project’s spatial approach, showing that location based analysis uncovers hidden inequities masked in state-level averages. 3.4 Intersectionality of Disability, Rurality, and Race O’Neil et al. (2021) explore how overlapping factors — such as rural location, race, and disability — amplify socioeconomic disadvantage. They call for disaggregated analyses to expose compounding inequities, reinforcing the importance of intersectional data modeling in the present study. 3.5 Policy and Accessibility Frameworks The U.S. Census Bureau (2024) outlines ACS methodology for tracking education, employment, and disability indicators, emphasizing the dataset’s reliability for social-equity research. Meanwhile, ADA-related policy reviews note the persistent need for evidence-based monitoring to ensure compliance and impact assessment.

3.6 Local Labor-Market Structures and Regional Disparities Johnson and Patel (2023) found that local labor market structures significantly moderate the relationship between educational attainment and income, particularly for disabled individuals in post industrial regions, reinforcing the need for geographically specific analyses. Summary of Gaps and Contribution Existing literature establishes that (1) education mitigates poverty overall, and (2) people with disabilities remain disproportionately poor. Yet, few studies combine these dimensions with geographic precision. This capstone addresses that gap by merging ACS tract-level data and applying data science methods to visualize where education’s protective effects are unequal. In doing so, it extends current scholarship through spatial modeling that links social policy with data-driven insights for targeted community interventions. 4 Methodology This project uses a quantitative, R-based approach to analyze how educational attainment and disability status jointly affect poverty across Illinois census tracts. All processing and analysis were completed in RStudio using the packages tidycensus, dplyr, tidyverse, janitor, ggplot2, sf, and related libraries. The workflow consisted of four main stages: data acquisition and aggregation, data quality assessment, feature engineering, and modeling. 4.1 Data Acquisition and Aggregation The data for this project were sourced from the American Community Survey (ACS) but used in two distinctly different ways. The primary modeling dataset was constructed exclusively from the ACS Public Use Microdata Sample (PUMS), which provides weighted person-level records containing detailed information on disability status, educational attainment, income, labor-force participation, and demographic characteristics. PUMS was selected as the modeling foundation because it allows direct examination of joint effects — such as disability × education or disability × race — that cannot be

178