CORS / Optimization Days 

HEC Montréal, May 29-31, 2023


HEC Montreal, 29 — 31 May 2023

Schedule Authors My Schedule

Queueing and Applied Probability SIG Best Student Paper Competition

May 30, 2023 03:30 PM – 05:10 PM

Location: St-Hubert (green)

Chaired by Yichuan Ding

3 Presentations

  • 03:30 PM - 03:55 PM

    Implications of Worker Classification in On-Demand Economy

    • Zhoupeng (Jack) Zhang, presenter, Rotman School of Management, University of Toronto
    • Ming Hu, Rotman School of Management, University of Toronto
    • Jianfu Wang, College of Business, City University of Hong Kong

    How should workers in the on-demand economy be classified? As contractors, employees, or somewhere in between? We study this policy question focusing primarily on the welfare of full-time workers, who have worked as much as employees but have been treated as contractors. We develop a game-theoretic queueing model with a service platform and two types of workers: full-timers who may choose gig jobs as primary income sources and are available for frequent participation on the platform, and part-timers who do gigs for supplemental incomes and have limited availability. We identify two issues with uniform classifications: when all workers previously treated as contractors are reclassified as employees according to rulings such as the Assembly Bill No. 5 in California, the profit-maximizing company may undercut (underpay or underhire) workers, and thus full-timers' welfare can decrease; when all are reclassified as ``contractors+," a UK practice and an intermediate status between contractor and employee that provides incomplete employee benefits but allows workers to self-join, workers may overjoin such that full-timers' utilization rate can remain low and their welfare will not be effectively enhanced. In light of these issues, we consider a differentiated scheme that classifies only full-timers as employees. This hybrid mode still suffers from undercutting but curbs overjoining; it may also do less harm to consumers and the platform operator than uniform classifications. We also study a differentiated dispatch policy that prioritizes full-timers over part-timers. We demonstrate the potential of this operational approach to counteract both undercutting and overjoining. Finally, we empirically calibrate the model and apply our insights to the ride-hailing market in California.

  • 03:55 PM - 04:20 PM

    Calculating service rates from empirically obtained state-dependent mean service times

    • Likang Ding, presenter, Mr.
    • Bora Kolfal, Prof.
    • Armann Ingolfsson,

    Queueing models are typically formulated in terms of service rates but recent empirical queueing research
    focuses on mean service times. We investigate the translation of empirical findings about workload-dependent mean
    service times into mathematical queueing models. To this end, we analyze service times in Markovian multi-server delay and loss queueing models with state-dependent service rates. Contrary to what is sometimes assumed, in general, mean service times are not the inverse of service rates, the rate of change in service rates is not always the opposite of the rate of change in mean service times, and service times in state-dependent Markovian systems are not exponentially distributed. We provide closed-form solutions to convert mean service times into service rates and vice versa. We find conditions under which monotonic mean service times imply monotonic service rates and vice versa. For loss systems, service times are standard finite-state-space phase-type random variables. For delay systems, we show that service times belong to a well-defined class of infinite-state-space phase-type random variables. These results enable empirical researchers to make more precise statements about the implications of results for state-dependent mean service times for state-dependent service rates.

  • 04:20 PM - 04:45 PM

    Potty Parity: Process Flexibility via Unisex Restroom

    • Setareh Farajollahzadeh, presenter,
    • Ming Hu, Columbia University

    This paper addresses the issue of unequal access to restrooms for women and LGBTQ+ individuals in buildings, commonly known as the "potty parity problem." We analytically investigate the optimal renovation plan for such buildings using three common fairness measures (totalitarian, Rawlsian, and strict equality). While an all-unisex restroom design may seem like a viable solution, our analysis shows that it can lead to a decline in all three fairness measures. Instead, we propose converting a men's restroom unit to a unisex restroom with some flexibility, improving all three fairness measures and reducing wait time disparities between gender-segregated restrooms. Thus, a little-flexible system outperforms a fully flexible one. We also explore additional benefits of unisex restrooms under different user behaviors and situations and present numerical results to support our findings.