Abstract
Policymakers have been slow to implement price-based congestion policies due in part to how little is known about the effects of policies that influence more than simply an individual’s commuting method. An individual can also alter her commute by choosing to travel from a different location. I develop a discrete choice structural model of the joint decisions of family residence and individual commuting modes, given the characteristics of the housing market and commuting options. I use rich individual-level data that allow me to include numerous unobserved heterogeneity terms; this strengthens the validity of my results relative to more aggregate analyses that are often undertaken. I am in the process of using model estimates to simulate the full set of effects of transportation policies that alter the financial and time costs of commuting. These policies include congestion pricing schemes, fuel or carbon taxes, and increased parking fees.
I estimate my model using individual-level Public Use Microdata Sample (PUMS) data from the 2005-2008 American Community Survey (ACS) for the Washington, D.C. metropolitan area. The PUMS data requires that I randomly assign individuals home and work locations, but the Census Bureau has granted me access to precise information on where individuals live and work from the restricted-access version of the ACS that I am currently using to improve the analysis. I augment the information in the ACS with data I have painstakingly assembled on the structure of the transportation network to map each individual’s optimal commute from each home and by each commuting method in the choice set. To do this, I use geographic information system (GIS) network analysis. The mappings allow me to create a unique dataset of individual commute options and characteristics that I use to estimate the trade-offs that individuals make among consumption, housing amenities, and leisure when choosing a home and commuting mode pair.
I also develop and plan to implement a methodology that (unlike previous literature) does not require that I treat groups of individuals living together as if they have a single set of preferences. Instead, I use a collective model of the household to account for the fact that spouses rarely commute to the same work location. This allows me to model the interplay between residential and commuting mode choices when spouses consider the proximity of their home to both work locations. I allow family members to have caring preferences, and I treat characteristics of the home as a family public good. The collective model requires observing individual consumption of at least one private good in the household to identify bargaining outcomes, and I use a novel assignable private good: the method and duration of each commute. This work is both an extension of the collective model to the residential choice and travel literatures as well as an application of the collective model to a problem
with discrete choices and a rich error structure.
JEL Codes: D13, Q52, R21, R41, and R48
Keywords: Residential Location, Travel Mode Choice, Intra-Household Allocation, Congestion Pricing, Discrete Choice Analysis, Geographic Information Systems
