paper12 complementarity between pricing and land use planning complementarity... · since 2004 has...
TRANSCRIPT
Complementarity between Pricing and Land Use Planning: Variations in Gasoline Price Elasticity of Transit Ridership in U.S.
Urbanized Areas
Bumsoo Lee,* Assistant Professor Department of Urban and Regional Planning University of Illinois at Urbana-Champaign Yongsung Lee, PhD Student School of City and Regional Planning Georgia Institute of Technology
Conditionally accepted to the Journal of the American Planning Association
*Corresponding author: Bumsoo Lee Assistant Professor Department of Urban and Regional Planning University of Illinois at Urbana-Champaign Champaign, IL 61820 [email protected]
2
Problem, research strategy, and findings:
A shift toward more sustainable transportation requires both adequate pricing of externalities from driving and supportive land use policies. However, a wide gap exists between advocates of pricing and land use planning approaches. Proponents of each approach underestimate the potential and role of the other approach and often ignore the complementarity and potential synergy between the two. To fill this gap, this study investigates the interaction effects between gasoline prices and land use (policy) variables using a panel data set of transit ridership in 67 urbanized areas between 2002 and 2010.
We found that pricing and land use planning are complementary in promoting transit patronage. Estimated elasticities of key variables show that while doubling the average gasoline price would increase transit ridership by 17.6% in an urbanized area with the mean population weighted density and no regional containment policy, a one standard deviation higher density and containment policy would respectively boost the gasoline price effect by 4.8% and 16.6%. These interaction effects are additions to the ridership premiums of density and policy themselves. The interaction effects were 4.7% and 16.4% when an urban compactness index was used instead of the density variable. Findings were consistent across several different types of analyses.
Takeaway for practice:
As the study shows, pricing schemes will be more effective in places where alternatives to automobility and supportive land use policies are exist. The influences of urban form on travel behavior will be strengthened when driving is adequately priced. Planners and policymakers should take advantage of the complementarity between pricing and land use planning approaches by implementing policies in combination and in well-coordinated ways.
Keywords: transit ridership, land use, gasoline price, elasticity, complementarity
Research support: NA
About the authors:
Bumsoo Lee ([email protected]) is an assistant professor in the Department of Urban and Regional Planning at the University of Illinois, Urbana-Champaign. His research interests include urban spatial structure, travel behavior, and economic impact analysis. Yongsung Lee ([email protected]) is a doctoral student in the School of City and Regional Planning at the Georgia Institute of Technology. His research focuses on land use and transportation interaction, polycentric urban form, and public transit.
3
Introduction
Passenger transportation accounts for about 70 percent of U.S. greenhouse gas (GHG) emissions in the
transportation sector (one-third of total GHGs). Thus, reducing vehicle miles traveled (VMT) by shifting
to low-carbon transportation modes such as mass transit and non-motorized travel is critical to meet
GHG reduction goals (Chapman, 2007; Banister, 2008; Anable et al., 2012). Beyond environmental
benefits, public transit offers many additional benefits, including congestion reduction (Nelson et al.,
2007), enhanced mobility for low-income and disadvantaged individuals (Deka, 2002; Giuliano, 2005;
Lubin & Deka, 2012), and various health benefits from active travel typically accompanying transit use
(Litman, 2010; Lachapelle et al., 2011; Morency et al., 2011).
However, public transit has been losing market share to automobiles over the second half of the
last century. Even the substantial increase in public subsidies for and investments in public
transportation since the early 1990s have not appeared to have a major impact on transit ridership. Only
since 2004 has transit ridership begun to grow faster than both population and highway VMT, allegedly
due to an unprecedented increase in gasoline prices (American Public Transportation Association, 2012).
Recent ridership growth highlights the role of prices in transportation markets (Figure 1) and has
prompted many studies of the gasoline price elasticity of transit ridership (Maley & Weinberger, 2009;
Haire & Machemehl, 2010; Yanmaz-Tuzel & Ozbay, 2010; Stover & Bae, 2011; Lane, 2012). The
present research focuses on the interaction between price and land use planning. Specifically, it
investigates how compact urban form and urban containment policy reinforce the price effect to promote
transit patronage.
[Figure 1 about here]
Although a shift toward more sustainable transportation requires both correct pricing for driving
and transit-supportive urban forms, there remains a gap between advocates of pricing measures and
4
proponents of land use policies. Advocates of each approach tend to underestimate the role of the other.
In particular, skepticism remains about the potential of a more sustainable urban form in reducing VMT
and carbon emissions (Moore et al., 2010; Echenique et al., 2012). Skeptics of recent planning initiatives
in line with smart growth principles typically emphasize the primacy of policies devoted to “getting the
price right.” However, land use planning and pricing approaches can be complementary and synergistic
rather than competitive and conflicting (Boarnet, 2010; Guo et al., 2011). Policy analysts and decision
makers should understand the complex interactions among various policy instruments to mitigate policy
conflicts and maximize synergistic effects (Stepp et al., 2009). Nonetheless, little empirical research has
been done on potential synergies among alternative approaches in transportation planning, pricing and
land use planning.
This study investigates the interaction effects of gasoline prices and urban form variables on
promotion of transit ridership in U.S. urbanized areas over a 108-month period from 2002 to 2010. It
employs a two-stage least square (2 SLS) regression analysis of transit ridership to take into account the
endogeneity of transit service supply. The interactions between gasoline price and urban form variables
are tested by including multiplicative interaction terms. Results show that transit ridership increases with
gasoline price and that the price effect is larger in those urban areas that have a more transit-supportive
urban form indicated by a higher population density or compactness index. The presence of a regional
level containment policy across jurisdictional boundaries is also associated with a larger elasticity of
transit ridership with respect to gasoline price. Policy implications of the findings are discussed in the
last section.
5
Land Use versus Pricing Approaches to Sustainable Transportation
The debate between advocates and skeptics of land use planning approaches to sustainable
transportation has been one of the most contested in recent years. Central to the debate is a question
about the role of sustainable urban form in reducing VMT and mitigating climate change (Cervero &
Landis, 1995; Giuliano, 1995; Moore et al., 2010; Winkelman et al., 2010). The debate is intellectually
rooted in profound gaps between planners and economists in their beliefs regarding expected functions
and effects of different types of market interventions (Ewing, 1997; Gordon & Richardson, 1997;
Brueckner, 2000; Knaap, 2008). On the one hand, the debate has contributed to expanding our
knowledge about the connections between land use and travel by stimulating rigorous research; on the
other, it may carry a counterproductive message to practitioners and policy makers that land use and
pricing approaches are competing rather than complement each other.
Proponents of land use approach emphasize the role of smart growth policies in mitigating
climate change (Ewing et al., 2008a; Ewing et al., 2008b; Marshall, 2008; Winkelman et al., 2010; Dulal
et al., 2011). This approach aims to reduce the demand for driving and make alternatives to driving more
attractive (Handy, 2006; Salon et al., 2012) through a wide range of land use planning strategies that
improve development density, especially near transit stops; land use mix and jobs-housing balance;
regional accessibility; and network connectivity (Cervero, 2003; Cervero & Duncan, 2006; Salon et al.,
2012). These shifts toward a more sustainable urban form are believed to achieve, beyond their
environmental benefits, the additional benefit of creating more diverse, healthy, and livable communities
(Levine et al., 2005; Heath et al., 2006; Levine & Frank, 2007; Aytur et al., 2008; Boarnet, 2011). Most
professional planners and local public officials subscribe to this view.
Skeptics of smart growth policies challenge land use strategies on two primary grounds:
feasibility and cost (Litman & Steele, 2012). Although an increasing body of literature has cumulated
6
evidence of significant links between compact urban developments and carbon-efficient lifestyles,
including less driving (Ewing & Cervero, 2001; Kuzmyak et al., 2003; Cao et al., 2009; Ewing &
Cervero, 2010), disagreement remains about the magnitude of urban form effects (Brownstone, 2008;
Transportation Research Board, 2009; Ewing et al., 2011; Heres-Del-Valle & Niemeier, 2011). Critics
of urban form strategies claim that the impacts are small compared with those of pricing and technology-
based policy measures (Brueckner, 2007; Staley, 2008; Mitchell et al., 2011; Echenique et al., 2012).
The weakening connection between land use and transportation has been attributed to ever-decreasing
transportation costs and the popularity of suburban lifestyles (Giuliano, 1995). Proponents of the market
approach also assert that compact development costs society by compromising housing affordability
(Phillips & Goodstein, 2000; Dawkins & Nelson, 2002) and worsening congestion (Sorensen, 2009).
Some authors argue that the welfare costs of anti-sprawl policies that may limit housing choices and
lifestyles are higher than policy benefits (Moore et al., 2010; Echenique et al., 2012) although smart
growth policies are likely to expand consumers’ choice sets in U.S. cities (Levine & Inam, 2004; Levine
et al., 2005; Handy et al., 2008).
Critics of land use planning approaches hold that correct pricing of transportation will be more
effective and efficient in inducing behavioral changes in travel (Anas & Rhee, 2006, 2007; Brueckner,
2007; Moore et al., 2010). In their view, many transportation problems—including excessive congestion,
pollution and GHG emissions—exist largely because individual drivers are not required to pay for the
negative externalities they impose on society (Small, 1997; Brueckner, 2005; Parry & Small, 2005).
Thus, pricing policies currently being implemented or under policy discussion, including congestion
pricing, carbon taxes, VMT taxes, and increased fuel taxes, aim to make driving relatively more
expensive and hence make alternative travel options more attractive although each of these pricing
instruments offers slightly different travel (dis)incentives (Greene & Plotkin, 2011).
7
Many planners believe that pricing measures alone cannot effectively reduce VMT (Winkelman
et al., 2009) because the demands for driving and transit use are inelastic with respect to fuel price
(Goodwin et al., 2004; Small & Van Dender, 2008; Litman, 2012). Graham and Glaister (2004) add that
simply stabilizing VMT at its present level would require a price increase greater than income growth
because travel demands are more sensitive to income changes than fuel price changes. A primary reason
for the price inelasticity of VMT is the lack of alternatives to automobility (Cervero & Landis, 1995;
Winkelman et al., 2009). Thus, it may be naive to expect significant changes in travel behavior to be
caused by better pricing alone, given that driving is the only realistic travel option in many urban areas
in the U.S. (Boarnet, 2010). Further, there is a concern that mobility-based congestion pricing may
further accelerate metropolitan decentralization and dispersion when it is not well coordinated with
proper land use strategies, resulting in deterioration of overall accessibility and eventually increasing
VMT (Levine & Garb, 2002). Lastly and perhaps most importantly, an equity concern that price
increases will disproportionately constrain the mobility of low-income households often makes pricing
instruments like congestion pricing unpopular (Levinson, 2009).
In the recent search for effective policies to reduce VMT and mitigate GHG emissions in the
transportation sector, researchers began to emphasize the complementarity or potentially synergistic
effects of land use and pricing policies. Cervero (1998) explored a range of complementary land use
planning and pricing policy options to build a functional and sustainable “transit metropolis”. Shen
(1997) also emphasized the importance of combining various policy approaches to tackle ever
worsening transportation problems in mega cities like Shanghai, China. Now, there is an urgent need for
integrating different policy approaches to meet the great challenge of climate change mitigation (Handy,
2006; Boarnet, 2010, 2011; Guo et al., 2011). Stepp et al. (2009) indeed demonstrate, by applying
system dynamics tools, that the interaction of two or more GHG reduction policies can lead to either
8
synergy or conflict effects and that so-called rebound effects can be effectively mitigated by
implementing synergistic policies in an integrated manner.1 However, despite the importance of
potential synergy between land use planning and pricing approaches, only a handful of studies on this
issue have been done to date.
Rodier et al. (2002), using two types of land use and travel demand models, show that pricing
measures (higher fuel tax and parking surcharges) are complementary to transit investments and transit-
oriented developments (TOD) in reducing VMT and emissions in the Sacramento region. Another
simulation study (Lautso et al., 2004) on seven European cities explicitly attempts to identify synergistic
effects and shows that the impact of combined car pricing and public transit policies is larger than the
sum of the effects of individual policies. A recent review (Rodier, 2009) of simulation results including
scenario planning studies of U.S. metropolitan areas also shows that land use, transit, and pricing
policies are mutually supportive when implemented in combination. Nevertheless, even state of the art
land use change and travel demand models have limited capacity to simulate the complex interactions of
combined policy instruments and hence are likely to underestimate synergy effects (Rodier, 2009).
Further, the result of a simulation study is largely determined by behavioral assumptions and parameters,
which can be studied only by empirical research.
In a comparison of travel mode choice in Boston, MA and Hong Kong, China, Zhang (2004)
shows that land use policies influence travel mode choice more effectively when complemented by
various pricing measures. More recently, Guo et al. (2011) tests whether compact land uses and
congestion pricing are mutually supportive in reducing VMT, using data collected from a pilot mileage
fee program in Portland, OR. This study shows that VMT reduction from congestion pricing is greater in
traditional neighborhoods, where alternative travel options are available, than in low density suburban
neighborhoods. It also reports that density and land use mix explain more of the variation in household
9
VMT when the pilot program was implemented. However, it is difficult to generalize study findings
because, as the authors acknowledge, their sample size is small and the Portland metropolitan area is
unique in having an urban growth boundary. Clearly, more generalizable empirical research on the
interaction effects between planning and pricing approaches is necessary.
Research Approach
Research Hypotheses
The main hypothesis of this research is that land use planning and pricing approaches are
complementary in promoting alternative transportation modes. In particular, we hypothesize that the
cross-elasticity of transit ridership with respect to fuel price is greater in more compactly developed
urban areas. When gasoline prices increase, public transit and non-motorized travel can be reasonable
substitutes for driving in compact urban areas with high density, mixed land use, and better transit
accessibility. However, even when gasoline prices are very high, a shift from driving to transit riding
will not be easy in sprawling regions that are built around automobility. Thus, we expect greater transit
ridership gains from a gasoline price shock in urban areas that are more compact and are pursuing smart
growth policies more actively.
A second and potentially alternative hypothesis is that the effects of gasoline prices on transit
ridership are larger in urban areas with more choice riders, who can switch easily between driving and
transit, than in places with many dependent riders, who have no alternative but to travel by transit
(Stover & Bae, 2011). If this is the case, auto-centric urban areas may have larger gasoline price
elasticities, to the extent that choice riders outnumber dependent riders. A time-series analysis of
gasoline prices and transit ridership in 33 metropolitan areas (Lane, 2012) indeed shows relatively large
elasticities in traditionally thought auto-dependent cities in the Midwest and Sunbelt—for example,
10
Indianapolis, IN; Atlanta, GA; Dallas, TX; Houston, TX; and Phoenix, AZ. The two hypotheses are not
necessarily mutually exclusive, and both are tested in the current research.
Research Methods
To test the hypotheses discussed above, this study examines how public transit ridership has
responded to gasoline price changes during the 2000s in the 67 largest urbanized areas (UAs) in the
United States. Our basic modeling strategy is a two-stage least squares (2 SLS) regression analysis with
multiplicative interaction terms between gasoline price and urban form variables. We employ the 2 SLS
regression analysis to take into account the endogeneity of transit service supply, reflecting the two-way
causal relationship between transit service supply and consumption. While the level of transit services,
including coverage and frequency, is an important determinant of transit ridership, the dependent
variable of our analysis, on the one hand, transit operators typically decide transit service level in
response to transit demand (ridership) on the other hand (Taylor et al., 2009). Thus, ordinary least
squares (OLS) regression models produce biased and inconsistent estimates of key coefficients because
of the two-way causal relationship.
As a solution, drawing from Taylor et al. (2009), we take a 2 SLS regression approach. We use
operating subsidies per capita from federal, state, and local governments to transit operators and UA
population as instrument variables (IVs) to predict transit vehicle revenue miles in the first-stage
regression model and use the predicted service supply variable as a key predictor of transit ridership in
the second stage. Ideally, IVs should be partially correlated with endogenous variables, transit service
supply in this study, but should not be correlated with the error term in the OLS model. Taylor et al.
(2009) use the percent of the UA population who voted for the Democrat in the 2000 presidential
election as an IV, expecting democratic-leaning UAs are more inclined to fund public transportation
11
systems. We choose to use transit subsidy per 1,000 people as an IV which has more direct influence on
the level of transit services given population size.
First stage:
Sit = f (POPit, SBit, Pit, GPit, LUit, GPit × LUit, Xit, Tt, Mt) (1)
Second stage:
TRit = f ( it, Pit, GPit, LUit, GPit × LUit, Xit, Tt, Mt) (2)
, where Sit is transit service supply in UA i and in month t; TRit is transit ridership; POP is UA population size; SB is operation subsidy; S is predicted transit service supply; P is transit fare; GP is gasoline price; LU is land use variables, such as population density, urban sprawl index, and urban containment policy indicator; X is the vector of other UA characteristics, including percent college and graduate students, freeway lane miles, and unemployment rate; T is the vector of time-specific variables, such as trend and post-gasoline-price-peak period indicator; M is the vector of 11 month dummies;
As shown in equation (2), we test the complementary effects between fuel price and land use
planning variables by examining the coefficients of interaction terms between the two groups of
variables (GPit × LUit). Multiplicative interaction models are popular in testing a conditional hypothesis,
which typically assumes that X affects Y depending on the condition of Z (Wright, 1976; Friedrich,
1982; Brambor et al., 2006). A positive coefficient of the interaction term in our analysis will
corroborate the main research hypothesis: that increase in gasoline prices has a greater stimulating effect
on transit patronage in urban areas that have more transit-friendly land uses and policies. All interaction
variables are used after centering (subtracting the mean) so that we can interpret the coefficients of base
terms—gasoline price and land use—as the impact of each variable when the other variable takes on its
mean value.
This research analyzes a unique panel data set covering the 67 UAs for 108 time periods, which
will produce more generalizable findings than previous cross-sectional or time-series studies. Because
12
longitudinal data may have area-specific and/or time-specific unobserved effects, OLS regression with
the pooled data (pooled-OLS) produces biased estimates. To control for the unobserved heterogeneity,
we employ a random effects model among available panel data analysis techniques. A fixed effects
model cannot be used because some of our key land use variables, the compactness index and
containment policy indicator, are time-invariant and hence are perfectly collinear with the fixed effects.
All regression models are estimated with heteroskedastcity consistent standard errors.
Variables and Data
The data set for this research covers the 67 largest urbanized areas (UAs) in the United States for the
108-month period between January 2002 and December 2010. Honolulu, HI; New Orleans, LA; Mission
Viejo, CA; and McAllen, TX, are excluded from the sample of UAs with more than 500,000 residents as
of 2000 for various reasons, such as missing values for key variables and disruptive events like
Hurricane Katrina. Table 1 summarizes descriptions and data sources of all dependent and independent
variables used in the analysis.
[Table 1 about here]
Natural logarithm of unlinked passenger travel (UPT) is used as the dependent variable of all
second-stage regression models. As a measure of transit service level, we use total vehicle revenue miles,
which we assume to be a function of UA population size, public subsidy for transit operation, and
seasonal effects. All transit-related data come from the National Transit Database (NTD), provided by
the Federal Transit Administration (FTA). We aggregate all transit-related variables across modes and
transit operators at the UA level. Monthly gasoline price data come from the consumer price index-
average price data, provided by the Bureau of Labor Statistics (BLS). Because the monthly average
price for unleaded regular gasoline is available only for 15 major metropolitan areas, average price by
13
city size class and census region has been used for other smaller urbanized areas. All nominal gasoline
prices are converted to 1983 real prices, using the consumer price index for all urban consumers (CPI-U).
UA population density, an urban compactness index, and a containment policy indicator are used
as measures of land use characteristics and policy. We use a population-weighted density measure:
While a conventional density measure simply divides population by total land area, this alternative
measure weights densities of small geographical units (in this study, census blocks) by population to
obtain the urbanized area level weighted density. This population-weighted density measure captures the
density that average residents of an urban area experience in their daily lives better than a conventional
density measure does (Transportation Research Board, 2009). Further, the population-weighted density
measure is less affected by the change in boundary definition of each urbanized area between census
years.2 We estimate population-weighted densities using 2000 and 2010 census data and linearly
interpolate densities for 108 months in the study period for each UA instead of using annual American
Community Survey (ACS) data, which are tabulated using the 2000 census UA boundary definition.
We also use an urban compactness index as an alternative variable to measure land use
characteristics beyond population density. Ewing et al. (2002) developed a four-factor sprawl index
based on 22 variables that measure residential density, land use mix, strength of activity centers, and
accessibility of street network for 83 metropolitan regions. We rename this index an “urban compactness
index” because a higher index value indicates a region with less sprawl. A dummy variable indicating
the presence of a “non-local” containment policy, which goes beyond local government jurisdictional
boundaries (Wassmer, 2006), is also included in combination with either of the two land use variables.
Other factors considered in this study include freeway land miles per 100 people, percent of
residents who are college and graduate school students, and unemployment rate. Some other
demographic variables, such as percentage of immigrants and population in poverty, were initially
14
considered but dropped after preliminary analyses. Monthly data are available for transit ridership,
service supply, gasoline price, and unemployment rate. However, annual data are used for freeway lane
miles and college students’ share because monthly data are not available. All independent variables
except dummy variables are used in natural logarithm so that we can interpret estimated coefficients as
transit ridership elasticity with respect to the corresponding variable.
Results
Elasticities of Transit Ridership
Table 2 presents the results of the 2 SLS regression analysis of the determinants of public transit
ridership. The first-stage regression analysis result is provided only for a pooled data model. The supply
of transit services measured in vehicle revenue miles is almost unit elastic with respect to the UA
population size (coefficient = 1.025). The pooled model result shows that a 10% increase in government
subsidies to transit operation per capita is associated with about 7% higher transit service level given
population size and, in turn, leads to about 8% additional transit passenger trips (0.6947 × 1.1697). This
combined elasticity of transit ridership with respect to operation subsidy is estimated to be smaller in a
fixed-effects model, 0.34.
[Table 2 about here]
Estimated cross-elasticities of transit patronage with respect to fuel price ranged from 0.245 in a
pooled data model to 0.082 in a random effects model. The results are comparable with elasticity
estimates from previous studies of U.S. transit systems, 0.08 – 0.42 (Mattson, 2008; Stover & Bae, 2011;
Litman, 2012). Whereas these estimates show that transit ridership is relatively inelastic to the change in
fuel prices, it should be noted that a large portion of the growth in public transit patronage in the U.S., a
15
12.9% increase between 2002 and 2008, can be attributed to gasoline prices, which doubled in real
dollars for the period.
Population density shows a moderately larger elasticity, 0.3 in a pooled data model and 0.15 in a
fixed-effects model. These estimates are larger than the weighted average elasticity drawn from 10
previous studies examined in a meta-analysis (Ewing & Cervero, 2010), 0.07. Since our analysis does
not control for urban design and land use mix variables, the population density variable is expected to
capture the effects of these other urban form characteristics to the extent that they are correlated with
density. An alternative land use variable, the urban compactness index, is also found to have a
significant stimulating effect on transit patronage, with the elasticity ranging from 0.185 in a pooled data
model to 0.72 in a random effects model. It is also found that a non-local containment policy has an
additional stimulative effect on transit use. The results imply that transit ridership is 7 to 36% higher in
an urbanized area with a containment policy than in an identical urban region without such a policy.
Because we have already controlled for population density or other urban form characteristics, this
ridership increase may not be a direct effect of urban containment policy. Rather, we interpret this
containment policy dummy as a catchall variable of various smart growth policies, recognizing that
those regions with non-local containment policies (such as Portland, OR; Seattle, WA; and Minneapolis-
St. Paul, MN) have also been pursuing other transit-supportive policies, such as transit-oriented
development.
The results of other control variables are consistent with our expectations, except for freeway
land miles per 100 residents. Transit fare per ride is significant with a negative sign in two models and a
larger college student share of adult population is associated with higher ridership across the board. The
higher the unemployment rate, the fewer commute trips and hence less transit patronage. Another
interesting finding is a 4 to 9% higher transit ridership during the period following gasoline price peak
16
($4.065 per gallon of unleaded regular gasoline) in June 2008, all else being equal. Although gasoline
prices have dropped after the peak, the unprecedented gasoline price shock in the 2000s seems to have
left a long-lasting, if not permanent, behavior change in the public transportation market. However, the
negative coefficient of the trend variable indicates that the underlying trend for the study period is still
downward. In other words, the uptick in transit ridership in the 2000s is more a result of the changes in
gasoline prices and other urban attributes than it is a long-term trend.
Variations in Gasoline Price Elasticity of Transit Ridership
Prior to examining the interaction effects between gasoline price and urban form variables, we conduct
an exploratory analysis to learn how the elasticity of transit ridership varies among urbanized areas with
regard to gasoline price. In this step, we estimate the elasticity in each UA by including interaction terms
between gasoline price and UA dummy variables in the pooled data model, without employing any
urban form or land use policy variables. The coefficient of gasoline price base term is, then, the
elasticity in the reference UA (Riverside-San Bernardino, CA), and we obtain the elasticities of all other
UAs by adding the coefficient of a corresponding UA interaction term to the reference elasticity.
Estimated elasticities range from -0.705 to 1.378 and are positive in 56 out of 67 UAs. The mean
and median elasticities of the sample are 0.284 and 0.262, respectively, falling within the range of
previous estimates. Estimated elasticities are plotted against the compactness index and the natural log
of per capita ridership, respectively, in Figure 2. The result confirms the main hypothesis: that cross
elasticity of transit ridership with respect to fuel price is greater in more compactly developed urbanized
areas such as Providence, RI; San Francisco, CA; and Portland, OR, than in places with more sprawl,
such as Riverside, CA; Atlanta, GA; and Raleigh, NC. The slope of the fitted line is about 0.002,
implying that a standard deviation higher compactness index is associated with about 0.05 point higher
17
gasoline price elasticity. The second hypothesis, the negative association between the elasticity and per
capita transit ridership, seems also supported, although that association is not strong. The results of this
exploratory analysis suggest that we need to control for the effects of the current level transit patronage
per capita to tease out the effects of urban form variables on gasoline price elasticity.
[Figure 2 about here]
Interaction Effects between Gasoline Price and Land Use Variables
Table 3 presents the results of our main models, in which land use and containment policy variables are
directly interacted with gasoline price. Per capita ridership is also interacted with gasoline price to
control for the significance of choice versus dependent riders. As explained above, the coefficient of
gasoline price presents the elasticity of transit ridership with respect to gasoline price when the
interacted urban form variable has the sample mean value and there is no non-local containment policy.
This main effect is estimated to range between 0.09 and 0.22, a slightly smaller effect than estimates
from models without interaction terms, shown in Table 2. The coefficients of interaction terms are
positive and significant in all models except for the fixed effects model.
[Table 3 about here]
These results indicate that compact urban form and high gasoline price are complementary in
promoting transit ridership. A higher gasoline price has more stimulative effects on transit patronage in
urban areas where higher density, mixed land use, better connected street networks, and more clustered
development make public transportation a more viable travel option. It is also suggested that policies
aiming to create a more sustainable urban form will have greater effects on transit ridership when
gasoline prices are higher.
18
Figures 3 and 4 summarize the magnitudes of the interaction effects. They show how gasoline
price elasticity changes depending on the level of density or compactness as well as the presence of
containment policy. Both graphs present the expected ridership changes in response to a 100% increase
in real gasoline price, projected by a study by the International Monetary Fund (Benes et al., 2012) to
occur over the next decade. Doubling gasoline price would increase transit patronage by 17.6% in an
urbanized area that has the mean population weighted density (10,964 persons per square mile) and no
regional level containment policy. However, ridership growth will increase to 22.4% and 25.7%,
respectively, in urban areas with higher densities by 1 and 2 standard deviations.3 It is also predicted that
adopting more aggressive land use and transportation policies, proxied by a regional containment policy,
would boost ridership gains by about 17% at all density levels. A similar analysis for the compactness
index is provided in Figure 4.
[Figures 3 and 4 about here]
It should be noted that the ridership gain premiums of a more compact urban form and policy
presented in Figures 3 and 4 present only interaction effects with pricing that occur in addition to their
own main effects. Elasticities of transit ridership with respect to density and the compactness index are
estimated at 0.26 and 0.17, respectively, in pooled models. Containment policy has an additional effect
of around 0.06.
Finally, we conducted an analysis with change variables to assess the robustness of the results
from analyses with level data discussed thus far. The dependent variable in models presented in Table 4
is the natural logarithm of ridership growth from the same month in the previous year, and most key
independent variables are also used in a form of logarithmic change. For the transit ridership per capita
variable, 1-year (12 periods) lagged value is used. We lose observations for 12 months in 2002 by
shifting to a growth model.
19
[Table 4 about here]
The results corroborate the findings that compact development and higher gasoline prices are
mutually supportive in promoting transit patronage. Interaction terms between gasoline price change and
urban form change variables have positive and significant coefficients, except for the fixed-effects
model. The increase in gasoline price is associated with transit ridership growth, and this association is
much stronger in urban areas where population density has also increased. The effect of gasoline price
increase is also found to be larger in more compactly developed urban areas. In accordance with the
analyses above, both the increase in transit ridership and the impact of gasoline price increase are larger
in urban areas where the level of per capital transit ridership was lower in the previous year. The results
of the ridership change model also show correct coefficient signs of transit fare and highway lane miles
variables: A decrease in fare and highway supply is associated with higher transit patronage.
Conclusions
A shift toward more sustainable transportation requires both adequate pricing of externalities
from driving and supportive land use policies. However, there is a wide gap between advocates of
pricing and land use planning approaches. Proponents of each approach underestimate the potential and
role of the other approach. While some authors increasingly call attention to the complementarity or
potentially synergetic effects of pricing and land use planning, little empirical research has been done to
support the idea. To fill this gap, this study investigated the interaction effects between gasoline prices
and land use (planning) variables using transit ridership data in 67 urbanized areas between 2002 and
2010.
The results suggest that pricing and land use planning are indeed complementary. Estimated
elasticities of key variables show that while doubling the average gasoline price would increase transit
20
patronage by 17.6% in an urbanized area with the mean density and no regional containment policy, a
one standard deviation higher density and a regional containment policy boost the gasoline price effect
by 4.8% and 16.6%, respectively. These interaction effects were 4.7% and 16.4% when an urban
compactness index replaced the density variable.
Although the present analysis is focusing on transit ridership and gasoline price change
controlled by market forces and not as a result of policy interventions, the findings are generalizable at
the policy level. Pricing policies such as fuel tax, carbon pricing, VMT tax, and congestion pricing are
all designed ultimately to raise the costs of driving, just as the increase in oil prices does, although the
incentive systems and geographical scope of such pricing instruments are different. Skeptics of smart
growth policies typically claim that pricing policies will have larger effects on drivers’ behavior than
will land use planning approaches. It may be true that, assuming political barriers are removed, pricing
can produce strong outcomes in the short run. However, the present analysis has elucidated that pricing
strategies will be significantly more effective in places where alternatives to automobility exist and
supportive land use policies are implemented. Therefore, pricing schemes should be implemented in
combination with land use policies and transportation investments attuned to smart growth principles to
achieve the policy objectives most effectively.
Planners should also readily embrace pricing policies. Concerns in planning communities remain
regarding potentially negative effects of aggressive pricing measures on the mobility of low-income
households. However, the equity issues associated with pricing schemes can be effectively addressed by
active recycling of revenues to improve public transportation and incentivize accessibility-enhancing
developments (Levinson, 2009). The Transit-Oriented Development Program in the Portland, OR,
metropolitan area is a good example of flexible use of transportation funds that improves housing
affordability as well as transit accessibility for low-income households. The present analysis also
21
suggests that transit-supportive urban form and land use policies will have much larger impacts under
conditions where gasoline prices are higher or driving is more expensive. The use of behavior
parameters that ignore potential interaction effects in modeling future land use and transportation is
likely to underestimate the impact of sustainable urban development, given that the real oil price is
anticipated to double over the next decade (Benes et al., 2012). Interaction effects and potential policy
synergies between various pricing and land use planning strategies warrant research priority in the future.
References
American Public Transportation Association. (2012). 2012 Public Transportation Fact Book. Washington, DC: American Public Transportation Association.
Anable, J., Brand, C., Tran, M., & Eyre, N. (2012). Modelling transport energy demand: A socio-technical approach. Energy Policy.
Anas, A., & Rhee, H.-J. (2006). Curbing excess sprawl with congestion tolls and urban boundaries. Regional Science and Urban Economics, 36(4), 510-541. doi: 10.1016/j.regsciurbeco.2006.03.003
Anas, A., & Rhee, H.-J. (2007). When are urban growth boundaries not second-best policies to congestion tolls? Journal of Urban Economics, 61(2), 263-286. doi: 10.1016/j.jue.2006.09.004
Aytur, S. A., Rodriguez, D. A., Evenson, K. R., & Catellier, D. J. (2008). Urban Containment Policies and Physical Activity: A Time–Series Analysis of Metropolitan Areas, 1990–2002. American Journal of Preventive Medicine, 34(4), 320-332. doi: 10.1016/j.amepre.2008.01.018
Banister, D. (2008). The sustainable mobility paradigm. Transport Policy, 15(2), 73-80. Benes, J., Chauvet, M., Kamenik, O., Kumhof, M., Laxton, D., Mursula, S., & Selody, J. (2012). The future of oil:
Geology versus technology IMF Working Paper 12-109: International Monetary Fund. Boarnet, M. G. (2010). Planning, climate change, and transportation: Thoughts on policy analysis. Transportation
research. Part A, Policy and practice, 44(8). Boarnet, M. G. (2011). A broader context for land use and travel behavior, and a research agenda. Journal of the
American Planning Association, 77(3), 197-213. Brambor, T., Clark, W. R., & Golder, M. (2006). Understanding interaction models: Improving empirical
analyses. Political analysis, 14(1), 63-82. Brownstone, D. (2008). Key relationships between the built environment and VMT. In T. R. Board (Ed.), Driving
and the Built Environment: The Effects of Compact Development on Motorized Travel, Energy use, and CO2 Emissions (Vol. SPECIAL REPORT 298, pp. 7): Transportation Research Board.
Brueckner, J. K. (2000). Urban sprawl: Diagnosis and remedies. International Regional Science Review, 23(2), 160-171.
Brueckner, J. K. (2005). Transport subsidies, system choice, and urban sprawl. Regional Science and Urban Economics, 35(6), 715-733. doi: 10.1016/j.regsciurbeco.2005.01.001
Brueckner, J. K. (2007). Urban growth boundaries: An effective second-best remedy for unpriced traffic congestion? Journal of Housing Economics, 16(3-4), 263-273. doi: 10.1016/jjhe.2007.05.001
Cao, X. Y., Mokhtarian, P. L., & Handy, S. L. (2009). Examining the Impacts of Residential Self-Selection on Travel Behaviour: A Focus on Empirical Findings. Transport Reviews, 29(3), 359-395. doi: 10.1080/01441640802539195
Cervero, R. (1998). The transit metropolis: a global inquiry: Island press.
22
Cervero, R. (2003). Growing smart by linking transportation and land use: Perspectives from California. Built Environment, 39(1), 66-78.
Cervero, R., & Duncan, M. (2006). 'Which Reduces Vehicle Travel More: Jobs-Housing Balance or Retail-Housing Mixing? Journal of the American Planning Association, 72(4), 475-490.
Cervero, R., & Landis, J. (1995). The transportation-land use connection still matters. Access(7), 2-11. Chapman, L. (2007). Transport and climate change: a review. Journal of Transport Geography, 15(5), 354-367. Dawkins, C. J., & Nelson, A. C. (2002). Urban containment policies and housing prices: an international
comparison with implications for future research. Land Use Policy, 19(1), 1-12. doi: 10.1016/s0264-8377(01)00038-2
Deka, D. (2002). Transit availability and automobile ownership. Journal of Planning Education and Research, 21(3), 285-300. doi: 10.1177/0739456x0202100306
Dulal, H. B., Brodnig, G., & Onoriose, C. G. (2011). Climate change mitigation in the transport sector through urban planning: A review. Habitat International, 35(3), 494-500.
Echenique, M. H., Hargreaves, A. J., Mitchell, G., & Namdeo, A. (2012). Growing Cities Sustainably. Journal of the American Planning Association, 78(2), 121-137.
Ewing, R. (1997). Is Los Angeles-Style Sprawl Desirable? Journal of the American Planning Association, 63(1), 107-126. doi: 10.1080/01944369708975728
Ewing, R., Bartholomew, K., Winkelman, S., Walters, J., & Anderson, G. (2008a). Urban development and climate change. Journal of Urbanism, 1(3), 201-216.
Ewing, R., Bartholomew, K., Winkelman, S., Walters, J., & Chen, D. (2008b). Growing Cooler: The Evidence on Urban Development and Climate Change. Washington, D.C.: Urban Land Institute.
Ewing, R., & Cervero, R. (2001). Travel and the built environment - A synthesis. Transportation Research Record, 1780, 87-114.
Ewing, R., & Cervero, R. (2010). Travel and the built environment. Journal of the American Planning Association, 76(3).
Ewing, R., Nelson, A. C., Bartholomew, K., Emmi, P., & Appleyard, B. (2011). Response to Special Report 298 Driving and the built environment: the effects of compact development on motorized travel, energy use, and CO2 emissions. Journal of Urbanism, 4(1), 1-5.
Ewing, R., Pendall, R., & Chen, D. (2002). Measuring sprawl and its impact. Washington DC: Smart Growth America.
Friedrich, R. J. (1982). In Defense of Multiplicative Terms in Multiple Regression Equations. American Journal of Political Science, 26(4), 797-833. doi: 10.2307/2110973
Giuliano, G. (1995). The weakening transportation-land use connection. Access(6), 3-11. Giuliano, G. (2005). Low income, public transit, and mobility. Transportation Research Record: Journal of the
Transportation Research Board, 1927(-1), 63-70. Goodwin, P., Dargay, J., & Hanly, M. (2004). Elasticities of road traffic and fuel consumption with respect to
price and income: a review. Transport Reviews, 24(3), 275-292. Gordon, P., & Richardson, H. W. (1997). Are Compact Cities a Desirable Planning Goal? Journal of the
American Planning Association, 63(1), 95-106. doi: 10.1080/01944369708975727 Graham, D. J., & Glaister, S. (2004). Road Traffic Demand Elasticity Estimates: A Review. Transport Reviews,
24(3), 261-274. doi: 10.1080/0144164032000101193 Greene, D. L., & Plotkin, S. E. (2011). Reducing greenhouse gas emissions from US transportation: Pew Center
on Global Climate Change Washington, DC. Guo, Z., Agrawal, A. W., & Dill, J. (2011). Are Land Use Planning and Congestion Pricing Mutually Supportive?
Journal of the American Planning Association, 77(3), 232-250. Haire, A., & Machemehl, R. (2010). Regional and Modal Variability in Effects of Gasoline Prices on U.S. Transit
Ridership. Transportation Research Record: Journal of the Transportation Research Board, 2144(-1), 20-27. doi: 10.3141/2144-03
Handy, S. (2006). The Road Less Driven. Journal of the American Planning Association, 72(3), 274-278. doi: 10.1080/01944360608976749
23
Handy, S., Sallis, J. F., Weber, D., Maibach, E., & Hollander, M. (2008). Is support for traditionally designed communities growing? Evidence from two national surveys. Journal of the American Planning Association, 74(2), 209-221.
Heath, G., Brownson, R., Kruger, J., Miles, R., Powell, K., & Ramsey, L. (2006). The effectiveness of urban design and land use and transport policies and practices to increase physical activity: a systematic review. Journal of Physical Activity & Health, 3, 55.
Heres-Del-Valle, D., & Niemeier, D. (2011). CO2 emissions: Are land-use changes enough for California to reduce VMT? Specification of a two-part model with instrumental variables. Transportation Research Part B: Methodological, 45(1), 150-161.
Knaap, G.-J. (2008). The sprawl of economics: A response to Jan Brueckner. In G. C. Cornia & J. Riddell (Eds.), Toward a Vision of Land in 2015: International Perspectives. Cambridge, MA: Lincoln Institute of Land Policy.
Kuzmyak, J. R., Pratt, R. H., Douglas, G. B., & Spielberg, F. (2003). Land Use and Site Design-Traveler Response to Transportation System Changes.
Lachapelle, U., Frank, L., Saelens, B. E., Sallis, J. F., & Conway, T. L. (2011). Commuting by public transit and physical activity: where you live, where you work, and how you get there. Journal of Physical Activity & Health, 8 Suppl 1, S72-82.
Lane, B. W. (2012). A time-series analysis of gasoline prices and public transportation in US metropolitan areas. Journal of Transport Geography, 22(0), 221-235. doi: http://dx.doi.org/10.1016/j.jtrangeo.2011.10.006
Lautso, K., Spiekermann, K., Wegener, M., Sheppard, I., Steadman, P., Martino, A., . . . Gayda, S. (2004). PROPOLIS: planning and research of policies for land use and transport for increasing urban sustainability: the European Commission.
Levine, J., & Frank, L. (2007). Transportation and land-use preferences and residents’ neighborhood choices: the sufficiency of compact development in the Atlanta region. Transportation, 34(2), 255-274. doi: 10.1007/s11116-006-9104-6
Levine, J., & Garb, Y. (2002). Congestion pricing's conditional promise: promotion of accessibility or mobility? Transport Policy, 9(3), 179-188. doi: 10.1016/s0967-070x(02)00007-0
Levine, J., & Inam, A. (2004). The market for transportation-land use integration: Do developers want smarter growth than regulations allow? Transportation, 31(4), 409-427.
Levine, J., Inam, A., & Torng, G.-W. (2005). A Choice-Based Rationale for Land Use and Transportation Alternatives: Evidence from Boston and Atlanta. Journal of Planning Education and Research, 24(3), 317-330. doi: 10.1177/0739456x04267714
Levinson, D. (2009). Equity Effects of Road Pricing: A Review. Transport Reviews, 30(1), 33-57. doi: 10.1080/01441640903189304
Litman, T. (2010). Evaluating public transportation health benefits: Victoria Transport Policy Institute Victoria, British Columbia, Canada.
Litman, T. (2012). Understanding transport demands and elasticities: How prices and other factors affect travel behaviorInstitute. Victoria, British Columbia: Victoria Transport Policy Institute.
Litman, T., & Steele, R. (2012). Land use impacts on transport: How land use factors affect travel behavior. Victoria, Canada: Victoria Transport Policy Institute.
Lubin, A., & Deka, D. (2012). The Role of Public Transportation as a Job Access Mode: Lessons from a Survey of Persons with Disabilities in New Jersey.
Maley, D., & Weinberger, R. (2009). Rising Gas Price and Transit Ridership. Transportation Research Record: Journal of the Transportation Research Board, 2139(-1), 183-188. doi: 10.3141/2139-21
Marshall, J. D. (2008). Energy-efficient urban form. Environmental science & technology, 42(9), 3133-3137. Mattson, J. W. (2008). Effects of rising gas prices on bus ridership for small urban and rural transit systems.
Upper Great Plains Transportation Institute, North Dakota State University. May, A. D., Kelly, C., & Shepherd, S. (2006). The principles of integration in urban transport strategies.
Transport Policy, 13(4), 319-327.
24
Mitchell, G., Hargreaves, A., Namdeo, A., & Echenique, M. (2011). Land use, transport, and carbon futures: the impact of spatial form strategies in three UK urban regions. Environment and Planning A, 43(9), 2143-2163.
Moore, A. T., Staley, S. R., & Poole, R. W. (2010). The role of VMT reduction in meeting climate change policy goals. Transportation Research Part A: Policy and Practice, 44(8), 565-574.
Morency, C., Trépanier, M., & Demers, M. (2011). Walking to transit: An unexpected source of physical activity. Transport Policy, 18(6), 800-806. doi: 10.1016/j.tranpol.2011.03.010
Nelson, P., Baglino, A., Harrington, W., Safirova, E., & Lipman, A. (2007). Transit in Washington, DC: Current benefits and optimal level of provision. Journal of Urban Economics, 62(2), 231-251. doi: 10.1016/j.jue.2007.02.001
Parry, I., & Small, K. (2005). Does Britain or the United States have the right gasoline tax? The American Economic Review, 95(4), 1276-1289.
Phillips, J., & Goodstein, E. (2000). Growth management and housing prices: the case of Portland, Oregon. Contemporary Economic Policy, 18(3), 334-344. doi: 10.1111/j.1465-7287.2000.tb00030.x
Rodier, C. (2009). Review of International Modeling Literature. Transportation Research Record: Journal of the Transportation Research Board, 2132(-1), 1-12. doi: 10.3141/2132-01
Rodier, C. J., Johnston, R. A., & Abraham, J. E. (2002). Heuristic policy analysis of regional land use, transit, and travel pricing scenarios using two urban models. Transportation Research Part D: Transport and Environment, 7(4), 243-254. doi: 10.1016/s1361-9209(01)00022-0
Salon, D., Boarnet, M. G., Handy, S., Spears, S., & Tal, G. (2012). How do local actions affect VMT? A critical review of the empirical evidence. Transportation Research Part D: Transport and Environment, 17(7), 495-508.
Shen, Q. (1997). Urban transportation in Shanghai, China: problems and planning implications. International Journal of Urban and Regional Research, 21(4), 589-606. doi: 10.1111/1468-2427.00103
Small, K. A. (1997). Economics and urban transportation policy in the United States. Regional Science and Urban Economics, 27(6), 671-691.
Small, K. A., & Van Dender, K. (2008). Long run trends in transport demand, fuel price elasticities and implications of the oil outlook for transport policy (Vol. 139): Organization for Economic.
Sorensen, P. (2009). Moving Los Angeles. Access: The Magazine of the University of California Transportation Center, 35(Fall), 16-24.
Staley, S. R. (2008). Missing the forest through the trees? Comment on Reid Ewing and Fang Rong's "The impact of urban form on US residential energy use".
Stepp, M. D., Winebrake, J. J., Hawker, J. S., & Skerlos, S. J. (2009). Greenhouse gas mitigation policies and the transportation sector: The role of feedback effects on policy effectiveness. Energy Policy, 37(7), 2774-2787. doi: 10.1016/j.enpol.2009.03.013
Stover, V. W., & Bae, C. H. C. (2011). Impact of Gasoline Prices on Transit Ridership in Washington State. Transportation Research Record: Journal of the Transportation Research Board, 2217(-1), 11-18.
Taylor, B. D., Miller, D., Iseki, H., & Fink, C. (2009). Nature and/or nurture? Analyzing the determinants of transit ridership across US urbanized areas. Transportation Research Part A: Policy and Practice, 43(1), 60-77. doi: 10.1016/j.tra.2008.06.007
Timms, P., Woodham, K., Milne, D., Ramjerdi, F., Vold, A., Minken, H., . . . Jarvi, T. (2005). Deliverable D8: Analysis and assessment of the practical impacts of combinations of instruments in an urban context, SPECTRUM (Study of Policies regarding Economic instruments Complementing Transport Regulation and the Undertaking of physical Measures: Institute for Transport Studies.
Transportation Research Board. (2009). Driving and the Built Environment: The Effects of Compact Development on Motorized Travel, Enery Use, and CO2 Emissions. Washington, D.C.: National Research Council, National Academy of Sciences.
Wassmer, R. W. (2006). The influence of local urban containment policies and statewide growth management on the size of united states urban areas. Journal of Regional Science, 46(1), 25-65.
25
Winkelman, S., Bishins, A., & Kooshian, C. (2009). Cost-effective GHG reductions through smart growth & improved transportation choices: An economic case for investment of cap-and-trade revenues Transportation and Climate Change Program (pp. 1). Washington, D.C.: Center for Clean Air Policy.
Winkelman, S., Bishins, A., & Kooshian, C. (2010). Planning for economic and environmental resilience. Transportation Research Part A: Policy and Practice, 44(8), 575-586.
Wright, G. C., Jr. (1976). Linear Models for Evaluating Conditional Relationships. American Journal of Political Science, 20(2), 349-373. doi: 10.2307/2110649
Yanmaz-Tuzel, O., & Ozbay, K. (2010). Impacts of Gasoline Prices on New Jersey Transit Ridership. Transportation Research Record: Journal of the Transportation Research Board, 2144(-1), 52-61. doi: 10.3141/2144-07
Zhang, M. (2004). The role of land use in travel mode choice: Evidence from Boston and Hong Kong. Journal of the American Planning Association, 70(3), 344-360.
26
Tables and Figures
Table 1. Description of study variables.
Variable Description Data Source2)
Dependent Variable Transit ridership Natural logarithm of monthly unlinked passenger travel (UPT) NTD Independent Variables Transit service supply Natural logarithm of vehicle revenue miles NTD Public subsidy Natural logarithm of federal, state, and local subsidies for service
operation per 1,000 people in 1983 dollars NTD
Population Natural logarithm of urbanized area population Census Gasoline price1) Natural logarithm of monthly gasoline price in 1983 dollars CPI, BLS Population density 1) Natural logarithm of population density weighted by census block
level population Census
Compactness index1) Sprawl index based on residential density, land use mix, strength of activity centers, and accessibility of the street network
Ewing et al. (2002)
Containment policy dummy Dummy variable indicating the presence of a non-local urban containment policy: 1 for UAs with the policy; 0 for no policy
Wassmer (2006)
Freeway lane miles Natural logarithm of freeway lane miles per 100 people FHWA College students share Natural logarithm of share of residents who are college and graduate
school students Census & ACS
Unemployment rate Natural logarithm of unemployment rate LAUS, BLS Trend Continuous monthly series: 1 for Jan 2002, 2 for Feb 2002, … , and
108 for Dec 2010
Post-peak dummy
Dummy variable indicating the periods after the gasoline price peak in June 2008: 1 for June 2008 and after; 0 for months before the peak
Month dummies 11 dummy variables for each month: Reference= May Notes: 1) All interacted variables including ln (gasoline price) are used after centering for the ease of interpretation. 2) NTD is the National Transit Database, maintained by the Federal Transit Administration (FTA); CPI, BLS is the
Consumer Price Index managed by the Bureau of Labor Statistics; FHWA is the Highway Statistics Series published by the Federal Highway Administration; ACS is the American Community Survey by the Census Bureau; LAUS, BLS is the Local Area Unemployment Statistics maintained by the Bureau of Labor Statistics.
27
Table 2. Two-stage least squares regression analysis of transit ridership determinants.
Pooled Model Fixed effects model Random effects model
Estimate t Estimate t Estimate t Estimate t Estimate t
1st Stage Subsidy per 1000 0.6947 89.24 ***
UA Population 1.0252 216.18 ***
2nd Stage
Predicted VRM 1.1697 231.84*** 1.2079 292.80*** 0.7422 15.75 *** 0.7122 14.49***
Gasoline price 0.2450 7.30*** 0.2785 7.73*** 0.0964 7.02 *** 0.0817 5.94***
Population density 0.3026 21.14*** 0.1473 4.78 ***
Compactness index 0.1848 20.42*** 0.7197 2.51 **
Containment policy D 0.0683 7.23*** 0.0666 6.60*** 0.3604 1.70 *
Transit fare per ride ‐0.0166 ‐1.58 0.0185 1.61 ‐0.1602 ‐10.24 *** ‐0.1881 ‐11.53***
Fwy lane miles/100 ‐0.0123 ‐0.97 ‐0.1871 ‐12.98*** 0.1196 3.28 *** 0.0959 3.48***
% college students 0.5792 21.33*** 0.5645 16.33*** 0.1908 6.53 *** 0.0909 3.13***
Unemployment rate ‐0.0118 ‐0.66 ‐0.0497 ‐2.51 ** ‐0.0622 ‐6.47 *** ‐0.0733 ‐7.97***
Trend ‐0.0040 ‐10.52*** ‐0.0047 ‐11.52*** ‐0.0008 ‐3.79 *** ‐0.0007 ‐3.10***
Post peak D 0.0525 2.42 ** 0.0920 3.89*** 0.0366 3.72 *** 0.0467 4.71***
January ‐0.0064 ‐0.41 ‐0.0033 ‐0.18 0.0044 0.22 ‐0.0159 ‐2.33 ** ‐0.0166 ‐2.42 **
February ‐0.0689 ‐4.45 *** 0.0440 2.39 ** 0.0532 2.67*** 0.0045 0.61 0.0015 0.20
March 0.0299 1.95 * 0.0126 0.69 0.0175 0.88 0.0195 2.78 *** 0.0218 3.06***
April ‐0.0001 ‐0.01 0.0111 0.62 0.0128 0.66 0.0080 1.21 0.0089 1.36
June ‐0.0158 ‐1.03 ‐0.0268 ‐1.53 ‐0.0205 ‐1.06 ‐0.0307 ‐4.69 *** ‐0.0285 ‐4.40***
July ‐0.0177 ‐1.14 ‐0.0336 ‐1.88 * ‐0.0311 ‐1.60 ‐0.0413 ‐5.83 *** ‐0.0425 ‐5.92***
August 0.0096 0.62 ‐0.0173 ‐0.97 ‐0.0216 ‐1.13 ‐0.0162 ‐2.39 ** ‐0.0189 ‐2.84***
September ‐0.0262 ‐1.67 * 0.0737 4.07*** 0.0747 3.85*** 0.0548 7.75 *** 0.0554 8.08***
Octorber 0.0199 1.26 0.0739 4.07*** 0.0742 3.81*** 0.0654 9.13 *** 0.0678 9.89***
November ‐0.0479 ‐3.03 *** 0.0647 3.52*** 0.0678 3.44*** 0.0181 2.45 ** 0.0154 2.17 **
December ‐0.0253 ‐1.58 0.0017 0.09 0.0070 0.35 ‐0.0444 ‐6.16 *** ‐0.0462 ‐6.74***
66 UA fixed effects …
Constant ‐7.4794 ‐108.09 *** ‐5.6364 ‐44.27*** ‐3.5204 ‐23.57*** 3.8434 4.34 *** 1.4875 1.05
N 7197 7197 6369 7197 6369
R‐Squared 0.950 0.957 0.957 0.993 0.371 Notes: 1) The dependent variable of the first stage regression is ln(vehicle revenue miles) and the dependent of all second stage
regression models is ln(unlinked passenger travel). 2) Parameter estimates of 66 UA dummy variables in fixed effects models are suppressed for reasons of space. * significant at 10%, ** significant at 5%, *** significant at 1%.
28
Table 3. Analysis of transit ridership with interaction effects between gasoline price and land use.
Pooled models Fixed effects model Random effects model
Estimate t Estimate t Estimate t Estimate t
Predicted vehicle revenue miles 1.1762 245.49 *** 1.2144 331.03 *** 0.7038 14.91 *** 0.6859 13.72 ***
Gasoline price 0.1763 5.21 *** 0.2215 6.10 *** 0.0925 6.62 *** 0.0880 6.12 ***
Population density 0.2612 19.21 *** 0.1444 4.80 ***
× Gasoline price 0.1082 2.09 ** ‐0.0237 ‐0.94
Compactness index 0.1691 20.16 *** 0.7352 2.64 ***
× Gasoline price 0.2137 5.38 *** 0.0800 3.99 ***
Containment policy D 0.0631 6.59 *** 0.0601 5.91 *** 0.3554 1.62
× Gasoline price 0.1656 4.56 *** 0.1644 4.31 *** 0.0541 3.22 *** 0.0380 2.81 ***
Per capita ridership × Gas price ‐0.0912 ‐3.33 *** ‐0.0736 ‐3.53 *** ‐0.0317 ‐2.07 ** ‐0.0208 ‐2.17 **
Fare per person mile traveled 0.1819 16.40 *** 0.1999 17.35 *** ‐0.1511 ‐10.33 *** ‐0.1735 ‐11.73 ***
Freeway lane miles/100 people ‐0.0252 ‐2.07 ** ‐0.1564 ‐11.12 *** 0.1139 3.22 *** 0.0839 3.03 ***
Percent college students 0.5525 21.59 *** 0.5856 18.12 *** 0.2119 7.06 *** 0.1369 4.65 ***
Unemployment rate ‐0.0188 ‐1.06 ‐0.0450 ‐2.29 ** ‐0.0645 ‐6.66 *** ‐0.0630 ‐6.60 ***
Trend ‐0.0036 ‐9.45 *** ‐0.0044 ‐10.85 *** ‐0.0009 ‐4.41 *** ‐0.0009 ‐4.20 ***
Post peak D 0.0335 1.54 0.0688 2.95 *** 0.0355 3.59 *** 0.0408 4.01 ***
11 Month dummies … … … …
66 UA fixed effects …
Constant ‐3.2616 ‐29.52 *** ‐3.2808 ‐24.32 *** 5.5414 7.05 *** 4.8163 6.33 ***
N 7197 6369 7197 6369
R‐Square 0.959 0.960 0.993 0.371
Notes: 1) In the first stage regression, ln(vehicle revenue miles) was regressed on ln(population), ln(operation subsidy), and 11
month dummies. The results are not presented for reasons of space. 2) The dependent variable of all second stage regression analyses is ln(unlinked passenger travel). 3) Parameter estimates of 11 month dummies for all models and 66 UA dummies in the fixed effects model are suppressed
for reasons of space. * significant at 10%, ** significant at 5%, *** significant at 1%.
29
Table 4. Analysis of transit ridership change.
Pooled model Fixed effects model Random effects model
Estimate t Estimate t Estimate t Estimate t
Predicted ln(VRM change) 0.6131 9.21 *** 0.5338 8.63 *** 0.5806 7.02 *** 0.3615 3.27 ***
ln(gasoline price change) 0.0538 7.02 *** 0.0499 6.91 *** 0.0399 5.83 *** 0.0367 5.82 ***
ln(population density change) 0.2348 3.39 *** 0.6870 1.77 *
× ln(gasoline price change) 0.4750 1.74 * 0.2280 0.92
Compactness Index 0.0002 3.22 *** 0.0048 3.64 ***
× ln(gasoline price change) 0.0009 3.07 *** 0.0006 2.71 ***
D Containmnet policy 0.0157 4.19 *** 0.0130 3.75 *** 0.1169 1.27
Lagged ln(ridership per capita) ‐0.0119 ‐6.62 *** ‐0.0099 ‐5.42 *** ‐0.3561 ‐36.93 *** ‐0.3121 ‐11.22 ***
× ln(gasoline price change) ‐0.0274 ‐3.50 *** ‐0.0322 ‐4.09 *** ‐0.0113 ‐1.62 ‐0.0207 ‐3.24 ***
ln(fare/PMT change) ‐0.0767 ‐7.26 *** ‐0.0940 ‐9.45 *** ‐0.0817 ‐8.38 *** ‐0.0959 ‐8.77 ***
ln(freeway/100 people change) ‐0.0564 ‐2.44 ** ‐0.0513 ‐2.18 ** ‐0.1348 ‐6.08 *** ‐0.0871 ‐3.09 ***
Change in % college students 0.0093 2.96 *** 0.0036 1.22 0.0067 2.35 ** 0.0012 0.52
Change in % unemployment 66 UA fixed effects
‐0.0127 ‐10.44 *** ‐0.0109 ‐9.51 *** ‐0.0059…
‐5.22 *** ‐0.0057 ‐5.53 ***
Constant 0.0077 3.07 *** 0.0090 3.78 *** 0.7394 29.46 *** 0.1664 3.87 ***
No. Obs. 6393 5661 6393 5661
R‐square 0.086 0.089 0.285 0.220
Notes: 1) In the first stage regression, ln(vehicle revenue miles growth) was regressed on ln(population growth), ln(operation
subsidy growth). The results are not presented for reasons of space. 2) The dependent variable of all second stage regression analyses is ln(unlinked passenger travel growth). 3) Parameter estimates of 66 UA dummies in the fixed effects model are suppressed for reasons of space. * significant at 10%, ** significant at 5%, *** significant at 1%.
30
Data sources: U.S. city average unleaded regular gasoline prices from Consumer Price Index data, U.S. Bureau of Labor Statistics; U.S. total unlinked passenger trips from the National Transit Database, Federal Transit Administration. Figure 1. Trends in gasoline price and transit ridership in the U.S., 2002-2012.
$0.00
$0.20
$0.40
$0.60
$0.80
$1.00
$1.20
$1.40
$1.60
$1.80
$2.00
0
100
200
300
400
500
600
700
800
900
1,000
JAN
02
MA
Y02
SE
P02
JAN
03
MA
Y03
SE
P03
JAN
04
MA
Y04
SE
P04
JAN
05
MA
Y05
SE
P05
JAN
06
MA
Y06
SE
P06
JAN
07
MA
Y07
SE
P07
JAN
08
MA
Y08
SE
P08
JAN
09
MA
Y09
SE
P09
JAN
10
MA
Y10
SE
P10
JAN
11
MA
Y11
SE
P11
JAN
12
MA
Y12
SE
P12
Mill
ions
National Transit Ridership Gasoline Price per Gallon (1983 Real $)
31
Figure 2. Elasticities of transit ridership with regard to gasoline price in U.S. urbanized areas.
Elasticity
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Compactness
0 20 40 60 80 100 120 140 160 180
Akron
Albany
Albuquerque
Allentown
Atlanta
Austin
Baltimore Birmingham
Bridgeport Buffalo
Chicago
Cincinnati
Cleveland
Columbus
Dallas
Detroit Grand Rapids
Las Vegas
Los Angeles Memphis
Miami
New Haven New York
Oklahoma City
Phoenix
Portland Providence
Raleigh
Riverside
San Diego
San Francisco
Springfield
Tampa
Tucson
Tulsa
Elasticity
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Ln(per capita ridership)
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Akron
Albany
Albuquerque Allentown
Atlanta Austin
Baltimore Birmingham
Boston Bridgeport
Buffalo
Charlotte
Chicago
Cincinnati
Cleveland
Concord
Dallas
Dayton
Denver
Detroit
Hartford
Indianapolis
Las Vegas
Los Angeles
Memphis
Miami
Nashville
New Haven New York Oklahoma City
Omaha
Philadelphia
Phoenix
Portland
Raleigh
Richmond
Salt Lake City
San Diego San Francisco
San Jose
Sarasota Tucson
Tulsa
Virginia Beach
Washington
32
Figure 3. Transit ridership changes due to a 100% increase in fuel price by population density and
regional containment policy.
8.8%
17.6%
22.4%
25.7%25.4%
34.2%
39.0%
42.3%
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
‐1 stdev Mean +1 stdev +2 stdev
Rid
ersh
ip e
last
icit
y w
rt f
uel
pri
ce
Population density
Without containment policy
With containment policy
33
Figure 4. Transit ridership changes due to a 100% increase in fuel price by compactness index and
regional containment policy.
1. While complementarity and synergy effects in policy instruments can be defined distinctively, we use these terms loosely to denote any positive interactions of combined policies that reinforce the effects of one another (May et al., 2006). The complementarity in this research refers to the first two cases in the table below.
Interaction effects Definition
Synergy Additivity Decreasing returns to packaging Incompatibility
welfare gain [A + B] > welfare gain [A] + welfare gain [B] welfare gain [A + B] = welfare gain [A] + welfare gain [B] welfare gain [A + B] < welfare gain [A] + welfare gain [B] welfare gain [A + B] < welfare gain [A] and/or < welfare gain [B]
Sources: Timms et al. (2005), modified by the authors.
2. Between 2000 and 2010 census definitions, land areas of 67 urbanized areas in our sample expanded by 19% on average. Area expansion was particularly significant in Southern UAs such as Charlotte, NC-SC (70.5%), Austin, TX (64.4%), and Raleigh, NC (62.1%). These significant boundary expansions artificially lower the density when assessed by the conventional density measure, although the majority of population still live in areas of similar density.
3. Providence, RI-MA, and Fresno, CA, urbanized areas have population weighted densities that are close to the
sample mean (10,964 persons per square mile). Densities in Boston, MA-NH-RI, and Philadelphia, PA-NJ-DE-MD, are about one standard deviation (6,109 persons per square mile) above the mean. A change from the density level in Atlanta,
7.3%
16.0%
22.2%
26.9%
30.8%
23.8%
32.4%
38.6%
43.4%
47.3%
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
‐2 stdev ‐1 stdev Mean +1 stdev +2 stdev
Rid
ersh
ip e
last
icit
y w
rt f
uel
pri
ce
Compactness index
Without containment policy
With containment policy
34
GA, to that of Dallas-Fort Worth-Arlington, TX, or San Diego, CA, would also approximate the density increase by a one standard deviation.