Accommodating Immigration Status and Self Selection Effects in a Joint Model of Household Auto Ownership and Residential Location Choice

摘要

As the proportion of immigrants in the US population continues to rise, it is becomingincreasingly important to understand their residential location choices and travel behavior in thetravel modeling and transportation policy making arena. This paper presents a joint model ofresidential location choice and auto ownership that explicitly accounts for immigration status andlength of stay in the United States as explanatory variables. In addition, the joint modelaccommodates error correlations across the choice dimensions thus accounting for residentialself-selection effects that may arise from unobserved preferences. The model takes the form of abivariate multinomial probit (MNP) model and is estimated using the computationally tractablemaximum approximate composite marginal likelihood (MACML) approach on a San FranciscoBay Area subsample of the 2009 National Household Travel Survey (NHTS). Model estimationresults show that immigration and length of stay are significant explanatory variables in bothresidential location choice and auto ownership, with immigrants displaying assimilation effects,i.e., they increasingly resemble non-immigrant households as the length of stay increases. Evenafter controlling for immigration effects and including residential location choice as anexplanatory variable in the auto ownership model, it is found that there are significant selfselectioneffects that are likely to dampen estimates of the impacts of land use changes on travelbehavior in policy forecasts. The paper demonstrates the need to account for immigrationvariables and self-selection effects in transportation forecasting models that inform policydecisions.