Understanding Transit Ridership Demand for a Multi-Destination, Multimodal Transit Network in an American Metropolitan Area: Lessons for Increasing Choice Ridership While Maintaining Transit Dependent Ridership
There is a growing body of evidence, including earlier Mineta Transportation Institute-sponsored research, showing that multi-destination transit systems are far more effective in attracting passengers and more efficient in use of resources to carry each passenger than central business district (CBD)-focused systems. At the same time, however, evidence is beginning to show that multi-destination transit systems appeal largely to transit-dependent riders (also called captive riders), whose demand for transit service appears to be highly elastic with respect to the shortening of transit travel time between origin and destination. Given the interest in using transit investments to lure people from their automobiles in order to reduce greenhouse gas emissions and reduce congestion, it is imperative that the appeal of such systems to choice riders (also called discretionary riders) also be understood. However, this issue remains as yet relatively unexplored.
In this study, we examine the Atlanta region’s transit system, and we derive lessons that can be applied to transit systems elsewhere that would like to increase ridership among choice and transit-dependent riders by better serving increasingly dispersed travel destinations through a multi-destination transit network. Atlanta provides an opportunity to explore the consequences of a multi-destination transit network for bus patrons (largely transit-dependent riders) and rail patrons (who disproportionately illustrate choice rider characteristics). This study is an extension of earlier work by the authors on the determinants of transit ridership demand for an overwhelmingly transit-dependent rider population in Broward County, Florida, whose transit agency (Broward County Transit, BCT) operates a bus-only multi-destination transit system. Atlanta provides an opportunity to extend this work to a metropolitan area with a much larger, multimodal, multi-destination transit system (Metropolitan Atlanta Rapid Transit Authority, MARTA) and to explore differences in the determinants of transit rider demand for different groups of transit riders.
DATA SOURCES AND METHODOLOGY
The method used in this research is to specify and estimate several statistical models that predict bus and rail transit work trips (the dependent variable) from one part of a metropolitan area (traffic analysis zone or TAZ) to another. In other words, we develop statistical equations that allow us to explain the influence of different types of variables on transit ridership. Explanatory variables include describing demographic and land use characteristics in zones where trips begin and end, as well as those describing the general cost of making the trip in terms of travel time. Our resulting models fall within a category of models known as direct demand models. The models use travel time estimates from the Atlanta regional transportation demand model runs for 2002, but the models used in this study are not sub-models of the models used by the ARC.
In this study, we employ two sets of models. For one set of models, the dependent variable consists of transit users who identified themselves as “bus or trolley bus” riders in the 2000 Census Transportation Planning Package (CTPP). For the other set of models, the dependent variable consists of transit users who identified themselves as “subway or elevated” riders in the 2000 CTPP. Many respondents undoubtedly used a combination of bus and rail modes to complete their trips, but the 2000 CTPP did not give such transit users a box to check. Multimodal respondents were forced to identify themselves as either “bus or trolley bus” or “subway or elevated” riders. Therefore, we treat the former group as (self-identified) bus riders and the latter group as (self-identified) rail riders, although many riders in either category undoubtedly use multiple modes for their trips. The explanatory variables used in the models include socioeconomic variables from the 2000 CTPP, land use variables defined by the local metropolitan planning organization (MPO), and variables that measure transit service quality (broken into three components: in-vehicle, out-of-vehicle, and transfer time) obtained from the travel time skims of the regional travel demand model.
Bus riders were overwhelmingly transit-dependent riders, and rail riders included a disproportionate number of choice riders. By and large, rail riders tend to come from zones with high levels of vehicle access and bus riders from zones with low levels of vehicle access. The model results highlight important similarities as well as differences between the two rider groups. In terms of similarities, both bus and rail trips are produced in larger numbers in zones with higher populations and higher population densities, and attracted to destinations with larger numbers of jobs, but generally not areas with the highest densities of employment. Both bus and rail riders are also generally quite sensitive to in-vehicle travel time and transfer time.
In terms of differences between bus and rail riders, bus riders tend to come from zones with lower income, lower vehicle access (as noted above), and higher minority populations. While rail riders also disproportionately come from minority zones, they come from zones with high levels of vehicle access and the income variable is not significant, except in the cases of rail riders destined to more dispersed destinations, who tend to come from zones with lower incomes, but also relatively high levels of vehicle access. Bus riders do not place the same importance on out-of-vehicle travel time to transit as do rail riders, suggesting that bus stops are distributed in such a way that most patrons can easily access the stops to board a bus and then exit the vehicle to reach their final destination. Rail riders, on the other hand, do place a premium on out-of-vehicle travel time, suggesting that they have difficulty with access to the stations and/or reaching their final destinations. This is not surprising given the small number of rail stations and their spatial distribution relative to the patterns of population and employment in Atlanta.
The results for the land-use variables also reveal important differences between bus and rail riders as well as insights into the importance of transit-oriented development (TOD). Bus riders in Atlanta are not influenced by the presence of a transit-oriented development at either the origin or destination. The CBD does not emerge as a statistically significant destination for bus riders; indeed, lower density employment clusters emerge as important destinations for these riders. For rail riders, on the other hand, the CBD does emerge as an important travel destination, and two of Atlanta’s TODs (Midtown and North Avenue) emerge as important contributors to rail patronage, in excess of what would otherwise be predicted by the employment levels or densities of these zones.
Transit commuters who consider themselves bus riders seem to want a grid of routes connecting the region’s employment centers with faster, more direct, and more frequent service. Shelters, good pedestrian connections and other amenities at transfer points are also implied as being important to these largely transit-dependent riders. With such amenities, many more transit-dependent riders will use transit, presumably relying less on friends and relatives for chauffeured auto rides. Many of these riders appear to use trains to speedily move from one part of the region to the other, relying on buses at one or both ends of the trip, so good transfer connections between buses and trains will also increase ridership of transit-dependent riders.
Transit commuters who consider themselves rail riders, who primarily access transit by automobile, want trains to take them to major employment destinations, including the CBD and some TODs. Serving more of these riders, who are more likely to be choice riders than their bus rider counterparts, will require extending lines into job-rich corridors and developing stations and station environments in those corridors with those qualities typical of the TODs like North Avenue and Midtown. The more that can be done with a network of several regional rapid transit lines, the greater the number of choice riders using transit in the Atlanta region. If a transfer to a bus is required to complete the trip, the service will attract lower status workers who none-the-less will live in auto-oriented environments and will make use of autos to access the system. Are these choice riders, as well? The model results suggest that many of them are choice riders. Their numbers would increase in a more expansive regional network of regional rapid transit lines that had excellent bus transfers to jobs within one to two miles of stations.
A grid of local buses tied into such a regional rapid transit system would greatly increase the number of transit-dependent riders, as well, because it would enable them to reach additional employment opportunities that are presently difficult or impossible for them to reach by transit. These results derive from a study of Atlanta, Georgia, but given their consistency with lessons derived from other locales, they provide important policy guidance to transit agencies seeking to increase ridership by both rider groups.