Richard Yun | Searcharxiv

arXiv2026

Sticky Homelessness (Working Paper)

Homelessness in American cities is becoming an ever more prominent issue, but its causes remain contested, ranging from mental health and substance abuse to housing affordability and local labor markets. To shed light on this issue, I construct a novel MSA-level national panel of homelessness counts using data from the U.S. Department of Housing and Urban Development. Using a long-differencing regression specification with the changes in rent entered in piecewise linear form, I find that rent increases predict large increases in homelessness rates, but decreases have little to no effect. The same conclusions are reached when I use a quasi-differencing moment condition, assuming a multiplicative mean specification. Then, I propose a theoretical model of the low-end housing market that explains the asymmetry I find in the data. Finally, I outline an IV strategy that instruments rent changes with a Bartik instrument of predicted employment growth interacted with local housing-supply elasticities. My findings suggest that homelessness is a housing problem; however, because the response is sticky downward, effective policy must complement housing-market interventions with measures that address barriers faced by people experiencing homelessness.

econ.GN↗

arXiv2025

Exploring LLM Reasoning Through Controlled Prompt Variations

This study investigates the reasoning robustness of large language models (LLMs) on mathematical problem-solving tasks under systematically introduced input perturbations. Using the GSM8K dataset as a controlled testbed, we evaluate how well state-of-the-art models maintain logical consistency and correctness when confronted with four categories of prompt perturbations: irrelevant context, pathological instructions, factually relevant but non-essential context, and a combination of the latter two. Our experiments, conducted on thirteen open-source and closed-source LLMs, reveal that introducing irrelevant context within the model's context window significantly degrades performance, suggesting that distinguishing essential from extraneous details remains a pressing challenge. Surprisingly, performance regressions are relatively insensitive to the complexity of the reasoning task, as measured by the number of steps required, and are not strictly correlated with model size. Moreover, we observe that certain perturbations inadvertently trigger chain-of-thought-like reasoning behaviors, even without explicit prompting. Our findings highlight critical vulnerabilities in current LLMs and underscore the need for improved robustness against noisy, misleading, and contextually dense inputs, paving the way for more resilient and reliable reasoning in real-world applications.

cs.AI↗