International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
A Longitudinal Investigation of Residential Location: Fuzzy Logic-Based Choice Set Generation and Panel Location Choice Models
Mahmudur Rahman Fatmi, Muhammad Ahsanul Habib, Subeh Chowdhury

Last modified: 11 May 2015


Choice of residential location is a long-term household decision that affects the land use and urban environment. Residential location decision is directly related with short-term travel activities of individuals. It is evident that residential location choice models are key components to understand and predict the urban growth and travel behaviour evolving over time and space. In recent years, there is a growing interest in the development of micro simulation-based integrated land use and transport models to investigate and forecast the dynamics of land use and travel patterns. The decision of households’ residential location choice has a profound temporal dimension, since families move one place to another at various stages of their lives. Despite the longitudinal nature of residential location process, existing literature often ignores this temporal dynamics and associated historical effects in case of choice of new residences.

This study develops a unique two-tier modeling framework to investigate the residential location choice decision that accounts for longitudinal effects of households’ housing history. In the first stage, a longitudinal residential search model of the relocating households is developed using the fuzzy logic techniques. One of the unique features of the proposed search model is that it accommodates the influence of prior locations on generating the potential choice set for the subsequent location. In the second stage, the study develops a location choice model using the potential locations generated in the earlier stage. The study uses a panel-based random-parameters logit modeling framework to investigate the household location choices. The model addresses temporal, longitudinal dynamics by assuming correlated sequence of location of each household’s housing career. The use of random-parameters logit model allows capturing longitudinal effects due to repeated choices.  The paper develops the models using parcel data as the choice unit to represent the decision process at the finer-grained scale.

The models are developed utilizing retrospective survey data from the Household Mobility and Travel Survey (HMTS) conducted from September 2012 to April 2013 in Halifax, Nova Scotia, Canada. The survey collected information on housing history of the households including detailed information on dwelling characteristics of at least three of the last residential locations. Additionally, the survey collected information on employment records, auto ownership history, daily travel activities, and attitudinal information. . The second important data set used in this study is the parcel level data from Nova Scotia Property Database 2013 that includes parcel attributes of all the parcels in Nova Scotia. Additional data sources used in this study include census tabulations from Statistics Canada, land use data from the Halifax Regional Municipality, and other relevant GIS data from the Desktop Mapping Technologies Inc. (DMTI).

The explanatory variables considered to explain the residential location decision include, parcel attributes, accessibility measures to different activity points, land uses, and neighbourhood characteristics. A number of interaction variables between households’ socio demographics, life cycle events and parcel or neighbourhood attributes are used to understand the different sensitivity of households’ socio demographics to parcel and neighbourhood related attributes.  Among the neighborhood characteristics, population density, number of bedrooms, percentage of non-movers, average household income, monthly shelter cost, and percentage of households with shelter cost to income share are tested in the model. Among the interaction variables, life cycle events within two years of relocation are interacted with lot size, such as household size increase due to birth or member move in, household size decrease due to death or member move out. The final model includes variables on the basis of hypothesis confirmation along with reasonable statistical significance.

The model results suggest that life cycle events, parcel attributes, accessibility measures and neighbourhood characteristics are the most significant determinants in the case of location choice decision. For example, households with an increase in the household size due to birth of child or addition of member within the last two years of the relocation prefer larger lot size. In contrast, households with a decrease in the household size due to death or move out of member within last two years of the relocation prefer a smaller lot size. In the case of accessibility, households are found to be extremely sensitive to accessibility to school. Among the neighbourhood characteristics, households exhibit preference for stable neighbourhoods with a high ratio of non-movers and high median shelter cost.  One of the interesting findings is that households prefer to reside in neighbourhoods with high disposable incomes. Households with suburban preferences are more likely to reside in neighbourhoods with a higher percentage of single detached dwellings. On the other hand, households with urban preferences exhibit a negative relationship. Furthermore, significant heterogeneity exists among the sample households in terms of accessibility and neighborhood characteristics. Finally, this model is expected to be implemented within the micro simulation-based integrated Transport, Land use and Energy Modeling System (iTLE) for Halifax, Nova Scotia, Canada.

Conference registration is required in order to view papers.