Do Smart Growth Strategies Have a Role in Cu
ing Vehicle Miles Traveled? A Further Assessment Using Household Level Survey Data
The B.E. Journal of Economic
Analysis & Policy
Topics
Volume 12, Issue XXXXXXXXXXArticle 37
Do Smart Growth Strategies Have a Role in
Cu
ing Vehicle Miles Traveled? A Furthe
Assessment Using Household Level Survey
Data
Sudip Chattopadhyay∗ Emily Taylor†
∗San Francisco State University, XXXXXXXXXX
†San Francisco State University, XXXXXXXXXX
Recommended Citation
Sudip Chattopadhyay and Emily Taylor (2012) “Do Smart Growth Strategies Have a Role in Cu
-
ing Vehicle Miles Traveled? A Further Assessment Using Household Level Survey Data,” The B.E.
Journal of Economic Analysis & Policy: Vol. 12: Iss. 1 (Topics), Article 37.
DOI: XXXXXXXXXX/ XXXXXXXXXX
Copyright c©2012 De Gruyter. All rights reserved.
Do Smart Growth Strategies Have a Role in
Cu
ing Vehicle Miles Traveled? A Furthe
Assessment Using Household Level Survey
Data∗
Sudip Chattopadhyay and Emily Taylo
Abstract
This paper draws on McFadden’s location choice theory and incorporates households’ resi-
dential choice decisions as a hierarchical process in a structural travel demand model. The pa-
per argues that such an approach can effectively tackle the problems of self-selection and multi-
collinearity. Contrary to previous findings, empirical results based on OLS and 3SLS reveal that
travel demand is highly elastic to certain smart-growth features, if they are measured at different
spatial scales. The results are robust against alternative sequencing of the hierarchical choice pro-
cess. An analysis of the quantitative impact of a change in the smart-growth and fuel-tax policies
eveals significant returns under both policies. Finally, a simulation based on California suggests
that smart growth policies substantially reduce household travel demand.
KEYWORDS: transportation demand, land use policies, self-selection, multicollinearity, hierar-
chical choice theory, structural equations model, three stage least squares
∗An earlier version of the paper was presented in the Summer 2011 Association of Environmental
and Resource Economists (AERE) conference held in Seattle, Washington, during June 9 – 10,
2011. The authors would like to thank the editor and the two referees for their valuable comments
that improved the paper. The authors also thank the participants of the conference for helpful
feedback. Data for this research are all from published sources and the findings of this research
have no potential conflict with any agency.
1. Introduction
The role of “smart growth” policies on auto-dependent transportation demand has
een extensively investigated in the economics and transportation literatures.1
Smart growth policies specifically focus on redevelopment of u
an areas to
educe auto-dependence. Key features of redevelopment include policies to
develop neighborhoods with high residential/employment densities and help build
communities with adequate variety of housing and employment mix that can cater
to a wide range of income levels. Another important feature of redevelopment is
the availability of or access to public transit. Policies to achieve high density
levels and a transit oriented u
an design are mutually reinforcing; on the one
hand transit stations located in high density neighborhoods are accessible to more
people, efficient transit network in cities with higher density neighborhoods, on
the other hand, can reduce trip lengths and time, thereby providing a low cost
alternative to private transportation. The proponents of the concept of “new
u
anism” that emphasizes smart growth believe that such an approach to
edevelopment stimulates integration of neighborhoods and thereby reduces
automobile travel (Ewing and Cervero, 2010).
The key finding of the studies investigating the role of smart growth on
auto-dependence is that the land use policies that characterize smart growth have
only a modest role in reducing the demand for vehicle miles traveled (VMT). For
the choice of an appropriate policy instrument to cu
transportation-related
energy use, this finding is disconcerting because, if a non-price instrument, such
as an innovative land use policy, does not work, the only recourse is to use a
price-based instrument, such as fuel or congestion tax, which is often not
politically expedient.
It is widely recognized in the existing literature on VMT demand that
households choose their residential location and consequently, the u
an-form
features of the location, such as residential/job densities, access to public transit
facilities that characterize “smart growth” or the lack of it. This process of self-
selection causes endogeneity in the VMT demand model. The literature also
ecognizes that different u
an-form features may have different spatial extent of
influence on travel behavior (Boarnet and Sermiento, 1998; Guo and Bhat, 2007).
For example, a job center can be identified with a larger geographic region, such
as a county, compared to a neighborhood with a certain level of residential
density. Unfortunately, linking the two aforementioned aspects of the rational
housing choice behavior has not been explored in the VMT demand literature.
1 Interesting studies include Bento et al. (2003, 2005), Boarnet and Sarmiento (1998), Brownstone
and Golob (2009), Crane (2000), Guo and Bhat (2007), Bhat and Guo (2007), Vance and Hadel
(2007), Su XXXXXXXXXXEwing and Cervero XXXXXXXXXXpresent a meta-analysis of the impact of smart
growth on VMT demand based on a comprehensive set of studies.
1
Chattopadhyay and Taylor: Smart Growth and Vehicle Miles Traveled
Published by De Gruyter, 2012
Building on past studies, this paper presents a simple, but an innovative modeling
approach that draws in part on the theory of residential choice – originally
developed by McFadden (1978), which models households’ choice of residential
location as a hierarchical choice process.
Following McFadden (1978), this study incorporates households’
esidential choice process in the VMT demand model as a sequence of
hierarchical decisions in which households choose, in sequence, certain
commonly used u
an-form features of land use at varying spatial scales. 2 Using
the 2001 National Household Travel Survey data (NHTS-2001), the study
empirically shows that the OLS estimation of such a demand model can
effectively tackle the multicollinearity problem frequently faced in VMT demand
estimation and can yield statistically as well as quantitatively significant impacts
of u
an-form features on VMT demand. Also, the three-stage least squares
(3SLS) technique developed to incorporate the aforementioned choice process to
estimate the VMT demand shows further improvement over the OLS in
addressing the endogeneity problem and in yielding statistically and quantitatively
significant impacts. Moreover, the findings based on the 3SLS model are shown
to be robust to alternative sequencing of the hierarchical choice process. Finally,
a simulation exercise for California, based on the estimated 3SLS model suggests
that innovative land-use mixes can result in substantial benefits for the state.
The paper addresses the aforementioned policy question with a special
focus on California, since understanding the role of smart growth strategies in
California is important and highly policy relevant. In California, the
transportation sector cu
ently accounts for nearly 38% of the greenhouse gas
(GHG) emission. This figure is expected to rise, since VMT in California is
increasing at a rate that far outpaces its population growth. SB 375 – a regional
transportation and land use planning statute and The California Global Warming
Solutions Act, 2006 (AB 32) are two recent initiatives that are expected to set the
direction of GHG mitigation strategies for the future. While SB 375 has a direct
focus on identifying and implementing smart growth strategies to reduce
transportation-related GHG emissions, AB 32 aims to achieve this goal indirectly
through its innovative ca
on offset policies linked to the cap-and-trade program.
SB 375 requires each region in California to create a prefe
ed growth scenario
that will minimize GHG emissions, and then ties state transportation funds to
projects that conform to that prefe
ed growth scenario. Global Warming Act
(AB 32) requires that by 2020 the state's greenhouse gas emissions be reduced to
2 Based on the theory propounded by McFadden (1978), Quigley XXXXXXXXXXand later Chattopadhyay
(2000) model household residential choice in a hierarchical, probabilistic choice framework and
apply the nested logit model for valuing housing amenities.
2
The B.E. Journal of Economic Analysis & Policy, Vol. 12 [2012], Iss. 1 (Topics), Art. 37
1990 levels, a roughly 25% reduction compared to the projected emission
estimate under the business-as-usual scenario.3
The next section discusses the econometric challenges su
ounding VMT
demand modeling and explains the methodological steps adopted in this paper to
address those challenges. Section 3
iefly discusses the household residential
choice process and the econometric methodology to incorporate such a process in
the VMT demand model. Section 4 presents the data sources and definitions.
The results of the estimation are presented and analyzed in Section 5. A
simulation is presented in Section 6 to assess the potential for California’s “smart
growth” initiatives. Section 7 provides a sensitivity analysis of the VMT demand
model against alternative hierarchical residential choice processes. Finally,
Section 8 provides conclusion.
2. Econometric Issues of VMT Demand Modeling
Modeling VMT demand is fraught with a number of challenging econometric
issues su
ounding the use of the common u
an-form features that affect VMT
demand. Chief among them are the problems of endogeneity and collinearity.
While the former is caused by the fact that households self-select in to
communities with prefe
ed u
an-form features, the latter is due to the fact that
commonly used measures of u
an form tend to move together. Recent research
has provided guidance in designing innovative u
an-form measures to tackle
collinearity and disentangle their independent impact on VMT demand (Bento et
al., 2003, XXXXXXXXXXInstrumental variable approaches (Boarnet and Sarmiento, 1998;
Vance and Hadel, 2007), including three-stage least squares (3SLS) (Brownstone
and Golob, 2009) and joint choice models of location choice and vehicle
ownership (Bhat and Guo, 2007; Guo and Bhat, 2007) have proven to be
successful in effectively addressing the problem of endogeneity and in
consistently estimating the effect of the u
an-form features on VMT.
A major concern in all these studies is that the effect sizes of the key
u
an-form features are, although statistically significant in many instances, they
are quantitatively too small to suggest policy action. Ewing and Cervero (2010)
in their meta-analysis report that the weighted average of the magnitude of
elasticities of VMT with respect to population density, job density, and distance to
transit are 0.04, 0.00, and 0.05, respectively. These extremely low magnitudes of