Modeling Work-Arrangement, ICT, Activity-Travel and Residential Location: Development of an Integrated Microsimulation Framework

Abstract

Traditional activity-based travel demand models (ABMs) primarily focus on out-of-home, in-person activities, overlooking the influence of flexible work-arrangements and virtual activities. This thesis advances ABMs by integrating both physical and virtual activity spaces into the activity generation and scheduling process. It introduces a novel integrated transport, land-use, and emission (iTLE) framework that captures the interdependencies between work-arrangement choices, ICT adoption, residential location, and activity-travel behavior. Advanced micro-behavioral econometric models and microsimulation techniques are developed to estimate individual and household choices across long-term, medium-term, and short-term decision horizons. A mixed logit model is formulated to capture unobserved heterogeneity in work-arrangement choices, explicitly including hybrid work as a distinct alternative. A modified Multiple Discrete Continuous Extreme Value model is developed to jointly model activity participation and time-allocation across physical and virtual environments, introducing fixed satiation and estimated translation parameters to account for diminishing marginal utility. To estimate residential location selection, a two-step modeling approach is implemented, combining a Gaussian Mixture Model for location search that generates spatially non-contiguous choice sets considering multi-dimensional attributes, and a latent segmentation-based logit model for final dwelling selection. Dimensionality reduction is performed using Random Forest and Principal Component Analysis, and the expected maximum utility from the work-arrangement model is incorporated to reflect the influence of work location flexibility on residential location. The model identifies two distinct household segments with divergent residential preferences, revealing heterogeneity in housing decisions. The microsimulation system is implemented for the 100% population of Halifax Regional Municipality, Canada. Activity generation is modeled using a Markov Chain Monte Carlo process that reflects substitution, complementarity, and fragmentation between physical-virtual activity types. ICT-device ownership and internet access are modelled to constrain virtual participation. Feedback mechanisms operationalize bidirectional interactions between work-arrangement and both activity-travel and residential location choice. Calibration and validation are performed for the base year and multiple simulation years to ensure model reliability. Scenario analysis provides behavioral insights into the impacts of work-arrangement, ICT-adoption, and housing supply on activity-travel patterns and urban form. The model serves as a decision-support tool capable of evaluating strategies for managing congestion, reducing urban sprawl, and achieving emissions reduction targets.