Dissertation Defence: Modeling and Microsimulating Residential Relocation Decisions Within an Integrated Urban Model and Testing for the COVID-19 Pandemic

Muntahith Mehadil Orvin, supervised by Dr. Mahmudur Fatmi, will defend their dissertation titled “Modeling and Microsimulating Residential Relocation Decisions Within an Integrated Urban Model and Testing for the COVID-19 Pandemic” in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Civil Engineering.

An abstract for Muntahith Mehadil Orvin’s dissertation is included below.

Examinations are open to all members of the campus community as well as the general public. Registration is not required for in person defences.

---

ABSTRACT

This thesis microsimulates residential relocation decisions within an integrated urban model – i.e., STELARS and tests for housing prices during an unprecedented socio-economic event such as COVID-19 pandemic. STELARS is an agent-based (i.e., individuals and households) model, which simulates land use, transportation, and energy-related decisions of each agent for a region. This thesis focuses on the residential relocation module, conceptualizing it as a four-stage decision-making process of: i) decision to move (mobility), ii) location search, iii) housing price, and iv) location choice. Advanced econometric and machine learning-based models are developed to represent the behavioral dynamics of residential relocation decisions. For example, a joint model is developed to capture the inter-dependency of mobility and location choice decisions. In case of mobility, a competing hazard-based model is formulated to capture continuous time dynamics of stay duration at a location. For location choice, a latent segmentation-based logit model is developed that accommodates unobserved heterogeneity and correlated sequence of repeated choices occurring over the life-time of households. To examine the effects of location choice on mobility, logsum parameters are estimated from location choice model and fed as an exogenous input into mobility model. Location search and price components are developed using Gaussian Mixture and latent segmentation-based autoregressive models, respectively. These behavioral models are methodologically consistent with the simulation technique of the STELARS. For example, STELARS adopts an event-based hybrid of continuous and discrete time microsimulation technique. In case of event-based approach, for example, mobility decisions are conceptualized to be triggered by life-cycle events such as childbirth, which is established through empirical evidence. In case of hybrid simulation, mobility is simulated as a continuous time decision using hazard-based model. Subsequent decisions involving location search, housing price, and location choice are simulated at discrete time steps. STELARS is implemented for the entire population of Okanagan region from 2011-2021. To incorporate the behavioral change of housing market during pandemic, separate price models for pre-pandemic and pandemic periods are estimated and implemented. This thesis performs a multi-year validation exercise, and presents a spatio-temporal analysis of predicted housing prices and location choices for different socio-demographic groups before and during pandemic.