EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI-assisted geospatial causal inference for U.S. food-waste bans: Robustness, transportability, and policy readiness

Maha Moussa, H.M. Hamouda, Khater A.E. Gadd

Waste Management Bulletin · 2026-12 · 4 (4) · 100385

DOI: 10.1016/j.wmb.2026.100385 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Waste Management Bulletin, 2026-12. Verfasst von Maha Moussa und 2 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Qassim University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

State food-waste bans differ in statutory coverage, exemptions, implementation, and geographic context. We reanalyze the corrected October 30, 2025 archive for the first five U.S. state bans, separating source-state causal evidence from counterfactual robustness and cross-jurisdiction transportability. The archive contains 54,580 waste records and yields 619 state-year observations for 36 states during 1996–2019. Corrected synthetic-control method (SCM) estimates remain the primary benchmark. Massachusetts has an average treatment effect on the treated (ATT) of −10.44% and a nominal two-sided unique-policy-unit placebo p = 0.0435 ; its 2000 optimization reruns remain negative. This exploratory five-state result is not significant under a Bonferroni familywise correction. An independent, differently specified SCM gives −9.49%. Geographic donor restrictions, donor deletion, pseudo dates, ridge regression, and validation-selected matrix completion generally preserve the negative contrast. A reproducible 2010 Census/BLS audit of population density and two labor-market measures confirms that geographic coordinates alone do not characterize substantive similarity. Massachusetts remains negative under modest context restrictions, but strong context penalties can reverse the sign at the cost of severely degraded pre-policy fit. Missing infrastructure, income, and regulatory covariates preclude a complete transportability claim. Geographic-block machine learning also shows severe overlap failure, and the staggered comparison has an imperfect pre-policy diagnostic ( p = 0.026 ). Processor proximity alone is insufficient to explain the Massachusetts contrast. The evidence supports a relatively robust, assumption-dependent Massachusetts benchmark rather than a common transferable effect. We propose a multidomain policy-readiness framework and prospective validation design.

Förderung

  • Qassim University

Zitieren

APA

Moussa, M., Hamouda, H., & Gadd, K. A. (2026). AI-assisted geospatial causal inference for U.S. food-waste bans: Robustness, transportability, and policy readiness. Waste Management Bulletin, 4(4), 100385. https://doi.org/10.1016/j.wmb.2026.100385

Vancouver

Moussa M, Hamouda H, Gadd KA. AI-assisted geospatial causal inference for U.S. food-waste bans: Robustness, transportability, and policy readiness. Waste Management Bulletin. 2026;4(4):100385. doi:10.1016/j.wmb.2026.100385

BibTeX
@article{Moussa2026,
  author = {Moussa, Maha and Hamouda, H.M. and Gadd, Khater A.E.},
  title = {AI-assisted geospatial causal inference for U.S. food-waste bans: Robustness, transportability, and policy readiness},
  journal = {Waste Management Bulletin},
  year = {2026},
  volume = {4},
  number = {4},
  pages = {100385},
  doi = {10.1016/j.wmb.2026.100385}
}
Kostenlos starten →