M.S. Applied Data Science - Capstone Chronicles 2025

13

Overall, the results advance existing literature by clarifying the geographic contours of disability related economic disadvantage and highlighting the need for regionally tailored, disability-inclusive policy design. 6.1 Conclusion The central conclusion of this study is that disability fundamentally alters the economic returns to education, resulting in persistently higher poverty risk across all educational levels in Illinois. Across every analytical stage — descriptive mapping, linear regression, logistic classification, and Geographically Weighted Regression — the results consistently showed that higher educational attainment lowers poverty rates, but does not eliminate the substantial poverty penalty associated with disability. The interaction effects reinforced the primary research question: disability weakens education’s protective effect, with disabled adults experiencing higher poverty even when holding equivalent degrees. These patterns were strongest in rural and post-industrial regions, where weaker labor markets and limited accessibility infrastructure create structural barriers that education alone cannot overcome. A second conclusion is that solving disability-related poverty requires more than increasing educational attainment. The persistence of poverty gaps — even among individuals with advanced degrees — indicates that policies must address accessibility, labor-market discrimination, and local economic conditions. These technical results align with national research showing lower earnings and higher unemployment among disabled adults and extend prior work by identifying where the disparities are most severe. Data quality checks — including MOE ratio screening and multicollinearity diagnostics — support the reliability of these findings, while also highlighting the need for caution when interpreting results in small rural tracts. Overall, the evidence demonstrates that disability inclusive workforce programs, ADA-aligned accommodations, and region-specific economic development strategies are essential to translating educational gains into equitable economic outcomes.

6.2 Recommend Next Steps/Future Studies The findings indicate that disability consistently elevates poverty risk across all education levels, even in communities where educational attainment is otherwise protective. Given this structural pattern, future studies should incorporate additional layers of socioeconomic context such as access to healthcare, transportation, and industry composition to better isolate the mechanisms driving the reduced economic returns to education for disabled adults. Extending the modeling framework to include longitudinal ACS data would also allow examination of whether these disparities widen or narrow over time. From a methodological perspective, integrating mixed-effects models or Bayesian spatial models could strengthen uncertainty estimation, especially in rural tracts where margins of error were higher. Validation using administrative datasets — such as state workforce records or vocational rehabilitation outcomes — would further improve reliability beyond survey-based ACS signals. Next steps should also emphasize developing actionable tools that translate these results into policy relevant insights. A public, web-based dashboard could leverage the final model and GWR outputs to help state agencies identify high-risk communities and allocate workforce development resources accordingly. Future modeling iterations could incorporate Explainable AI (XAI) techniques, such as SHAP values, to better communicate drivers of poverty disparities to non-technical stakeholders. Finally, targeted qualitative research — such as interviews with community organizations or disability employment programs — would complement the quantitative findings and help contextualize the structural barriers revealed in the data. ACKNOWLEDGMENTS I would like to express my sincere appreciation to my instructor for the thoughtful guidance, methodological clarity, and detailed feedback provided throughout each milestone of this project. Their recommendations on data validation, model interpretation, and the integration of ACS and PUMS sources greatly

188