Sequential Request Rejection Patterns in Ride-Hailing Systems: A Zero-Inflated Request-Level Analysis
Sustainability (Switzerland), cilt.18, sa.18, 2026 (SCI-Expanded, SSCI, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 18 Sayı: 18
- Basım Tarihi: 2026
- Doi Numarası: 10.3390/su18189292
- Dergi Adı: Sustainability (Switzerland)
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), Scopus, CAB Abstracts, Geobase, INSPEC
- Anahtar Kelimeler: matching time, ride-hailing, sequential request rejection, zero-inflated models
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Ride-hailing research has largely examined driver responses as isolated binary accept–reject decisions, offering limited insight into how unsuccessful assignments accumulate at the passenger-request level and affect matching performance. This study examines sequential rejection exposure and matching time using one month of operational data from Istanbul. The analysis covers 236,384 requests that ultimately achieved successful driver–passenger matches; unmatched and canceled requests were unavailable. GIS-based free-flow benchmarks supported the delay measure. Rejection counts, characterized by overdispersion and 61.3% zeros, were compared across Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial (ZINB) specifications. Matching time was analyzed using a zero-inflated ordered probit model, with inference based on driver-clustered robust standard errors. The ZINB model provided the best overall fit. Each additional minute of pickup time and realized delay was associated with 15.4% and 4.6% higher conditional rejection counts, respectively, whereas longer realized trip times were associated with lower rejection exposure. Pickup time was also associated with a 4.687-percentage-point reduction in immediate matching probability. Pickup time, trip time, and delay describe the final successful match and were unavailable for previously rejecting drivers; the estimates therefore represent request-level associations conditional on eventual matching rather than causal effects on prior rejection decisions.