Saturday, April 4, 2009

Studies Analyzing the Influence of Urban Form on Transit Patronage

TCRP Report 16, published in 1996, presents results from Project H‐1, An Evaluation of the
Relationships between Transit and Urban Form. Report 16 consists of two volumes, each
containing two reports. The first volume includes a comprehensive literature review of studies
analyzing the relationship between transit and urban form for the period to 1995. The second
volume consists of a practitioner’s guidebook on patterns of development that encourage
transit patronage and on mode accessibility and catchment areas for rail transit.
The most important work prior to Report 16 is Pushkarev and Zupan (1977). This
publication presented “land‐use thresholds” at which different types of transit were feasible
investments. The methodology used single‐equation ordinary least square (OLS) regression
analysis. The choice of this method was dictated by the paucity of data available at the time as
well as the desire to present results as nomograms. A nomogram is a graph with which one can
find the value of a dependent variable given the values of two or more independent variables,
with only the use of a straightedge. The nomograms were designed to facilitate a planner’s
choice of a feasible transit alternative, given values of current or expected density levels and
other relevant variables.
The determinants of transit demand used by Pushkarev and Zupan were the size of the
central business district (CBD), measured in non‐residential floor space; the distance of a site
from the CBD; and residential density. The study also accounted for socio‐demographic
characteristics affecting transit patronage, such as vehicle ownership levels, household size and
income.
There are problems with this study as well as with a later one by Pushkarev and Zupan,
(1982). As part of this review, CUTR researchers tried to replicate numerical examples in the
study, but encountered several problems that made replication impossible. These problems
are similar to those of papers reviewed later. An important example of such a problem is the
lack of a formalized behavioral framework, a deficiency that in turn results in poorly specified
empirical equations.
In a subsequent update of their 1977 study, Pushkarev and Zupan (1982), examined the
feasibility of fixed guide‐way transit under the assumption that all work travel was to the CBD.
This assumption would be quite restrictive today, given the multi‐centered character of many
metropolitan regions.
Multicollinearity impairs the reliability of these estimates, as recognized by the authors
themselves. Also, lack of causality is a problem, for the estimated elasticities merely support a
direct relationship between transit patronage and population density. This causality problem,
which affects most findings in this research field, is discussed in a later section of this report.
Finally, the authors do not employ a model that accounts for inherent, unobserved regionspecific
characteristics that might affect the reliability of estimates. A fixed effect model
controlling for transit provider unobserved heterogeneity could provide a superior model.
Using Report 16 as a reference, Kuby et al. (2004) analyze the determinants of light rail
transit ridership with a multiple regression model using weekday boardings for 268 stations in
nine cities. For each city, five categories of independent variables accounting for land‐use and
other factors are used. The authors hypothesize that employment within walking distance of
each station is the most important factor for work trips. The model also controls for the
relevance of nearby airports and for city‐specific unobserved effects that might affect weekly
boarding, such as the presence of an international airport. The study finds that an increase of
100 persons employed within walking distance of a station increases boarding by 2.3
passengers per day while an increase of 100 persons residing within walking distance of a
station increases boardings by 9.2 passengers per day. The study also finds higher residential
population to be associated with higher weekly boardings and that the CBD variable is not
statistically significant, indicating that centrality is no longer relevant in determining light rail
ridership. This result could, however, be due to faulty test statistics produced by the high
correlation between the model’s measures of centrality and the CBD dummy.
Kuby et al. made some important improvements to the methodology of Report 16. First,
they captured the effect of the CBD on boardings by introducing a dummy variable for CBD
location. In contrast, Report 16 only examined ridership at non‐CBD stations. Second, Kuby et
al. included employment near non‐CBD stations. Report 16 included employment within the CBD, but it ignored employment around other stations. Third, they included accessibility to
non‐CBD stations. Report 16 computed distances from the stations to the CBD, but it ignored
stations’ accessibility to other stations. Finally, Kuby et al. used residential population within
CBD as an independent variable, while Report 16 did not.

report : Integrating Transit and Urban Form
page : 7 - 11

1 comments:

goraya said...
This comment has been removed by the author.