Abstract

Charlotte-Mecklenburg, in the midst of years of strong growth, has adopted a Centers and Corridors Vision to sustain its prosperity and check the potential loss of jobs and residents to adjacent suburban jurisdictions. Successful integration of land use and transit planning is needed to avoid choking gridlock and gradual stagnation. Recent development is extensive but typically at very low suburban densities. This is the second in three papers that show how an intensive six-month study assessed transit opportunities in five corridors and showed how to alter current land use trends to better support transit and the Centers and Corridors Vision. This paper focuses on the innovative approaches to ridership projections.

Introduction

Estimating the likely ridership impacts and economic benefits of coordinated transit and urban development has never been easy. For about a century, there has been little strategic co-development of new transit systems and land uses from which to draw empirical knowledge. In addition, few metropolitan areas have enough sophisticated and robust regional forecasting models to capture the often subtle relationship between land use and transit. This is especially true for medium-size metropolitan areas where most new fixed-guideway transit investments are being programmed. Such were the obstacles we, as consultants, faced in producing something of value to our clients.

Estimating Ridership Impacts of Alternative Land Use Scenarios

Charlotte-Mecklenburg's traditional four-step model, calibrated mainly in the 1960s, was clearly not up to the task of estimating the ridership implications of various transit-oriented development (TOD) scenarios. This is not a criticism but an acceptance that the traditional four-step model was never intended to produce fine-grained analyses of localized land-use impacts. The four-step model using data gathered at the level of traffic analysis zones (TAZs) provides a macro-scale framework for identifying where to target corridor-level improvements. However, TAZs usually encompass a much bigger geographical area than the half-mile ring associated with a TOD or transit village (approximately the distance covered in a 5-10 minute walk). There are also problems with model specification. For instance, the relationship between transit ridership and density is known to follow a negative exponential form -- as distance from a station increases, ridership rates initially fall rapidly but eventually the rate of decline tapers. Also, four-step models rarely explicitly include density as a predictor variable of either trip generation or modal splits. The inability of traditional models to capture the ridership implications of concentrated development around stations, subregional jobs-housing balance, and other features of contemporary built environments called for a supplemental, more heuristic modeling approach.

We essentially triangulated the analysis. The Charlotte-Mecklenburg DOT used their four-step model to generate Year 2025 ridership estimates. Population and employment projections produced by econometricians from the Wharton School formed the key input to DOT's ridership forecasts. To refine the estimates, we drew upon experiences of other North American cities that have invested in fixed-guideway systems over the past two decades. To do so, we relied on a database produced under Project H-1 of the Transit Cooperative Research Program (TCRP) of the National Research Council. The data base contains information on land use environments, transit service levels, and daily boardings for 314 station settings pooled across eight U.S. cities (plus Calgary, Canada) that have built recent-generation light rail transit (LRT) systems or extensions. Importantly, the database provided information on 1990 population densities with half-mile rings of stations, in addition to variables specifying the size and density of downtown employment. An implicit assumption was that relationships that held in places like San Diego, St. Louis, and Portland would be applicable to Charlotte-Mecklenburg. We felt this to be a reasonable assumption, and our clients concurred.

The supplemental modeling served two key purposes. As noted, it provided a more credible basis for testing land-use scenarios. Equally important was its providing a crosscheck of ridership projections from two independent analyses based on different modeling structures and data inputs. To the degree the estimates from independent modeling approaches were consistent for baseline scenarios, we would have all the more confidence in the reliability of projections.

Model Refinements and Estimates

We quickly realized that the estimated model from the TCRP H-1 project suffered from specification problems. We therefore, reexamined the original data and re-estimated a model. Most worrisome, the TCRP H-1 model failed to include a variable measuring transit service intensity. This was a serious omission, for it meant that the impacts of the variables associated with service levels that are included in the TCRP H-1 model would be overstated. Specifically, since denser areas are typically rewarded with more frequent transit services, the TCRP H-1 model likely exaggerated density's impact on ridership. City-specific (dummy variable) controls were also introduced to statistically capture the unique effects of each city on ridership rates. Not only did these refinements produce a much better fitting model, but most importantly a more valid relationship -- more consistent with recent research -- was captured between density and ridership. Whereas TCRP H-1 estimated the elasticity between ridership and residential densities to be 0.591, we produced a far more conservative estimate of 0.197 -- that is, every 10 percent increase in residential density is associated with a 1.97 percent increase in transit boardings, ceteris paribus.

Several other refinements were necessary. Since TCRP H-1 contained data only for LRT, we had to invoke assumptions regarding the likely impacts of bus rapid transit (BRT). It was assumed that as dedicated, fixed-guideway services, BRT would exhibit many of the same service features as rail along line-haul portions of trips. Thus, only variables reflecting differences in feeder connections (e.g., park-and-ride provisions; feeder intensities) were adjusted. Based on experiences in Ottawa and Pittsburgh, it was also assumed that the "service intensity/quality" of BRT operating on 5-minute peak headways would be comparable to a three-unit rail train service operating on 10-minute peak headways.

Lastly, we post-processed the results to account for impacts of concentrated suburban job growth since TCRP H-1 only contained data on downtown employment size and density. Based on experiences from five rail areas in California, metropolitan Washington, D.C., and Edmonton, it was conservatively assumed that 9 percent of commute trips made by employees working in station areas outside of downtown would be by fixed-guideway transit. Forecasts from the modified TCRP H-1 data base were generally higher than from Charlotte's own four-step model at the corridor level. The independent projections were reviewed by a panel of experts on a station-by-station and corridor-by-corridor level and final base-line estimates were agreed upon. Where the TCRP H-1 data became most useful was in pivoting off the base-line "trend" projections to capture impacts of land-use scenarios.

Presenting model results in a way that would help Charlotte-Mecklenburg planners in their day-to-day work was essential. Two key outputs were produced. One was a series of graphs for each of the five corridors that showed how total ridership and cost per passenger would vary as population and employment densities increased between trend-line and maximum build-out scenarios. On the North line, for instance, it was estimated that raising residential densities by 140 percent (above trend estimates) would increase ridership by 10 percent; comparable rates of employment density increases would likely raise ridership by 41 percent. We also benchmarked the amount of station-area growth that would be necessary along each of the five corridors to achieve cost-effectiveness targets, defined as no more than $1.50 total cost per passenger (in today's dollars) for rail services and $1.00 for busway services. The clear message is that pro-active planning will be essential in concentrating enough housing and employment development around station to make fixed-guideway transit investments cost-effective.

Estimating Economic Benefits

Refined ridership forecasts formed the basis for estimating economic benefits. Only benefits which could be reliably monetized were quantified -- reduced traffic congestion, reduced vehicle motoring costs, reduced accidents, and reduced external costs (e.g., pollution). Other benefits, like improved job accessibility for inner-city residents, were treated qualitatively.

Benefits were compared to costs for the Year 2025. A differential analysis wherein network travel times, accident levels, etc. were compared with and without fixed guideway investments yielded these benefit estimates. Year 2025 model network outputs on total highway vehicle miles traveled (VMT), adjusted based on the supplemental analysis, were relied upon for scaling benefits. Unit costs for each component (e.g., external costs) expressed on a vehicle mile basis (adjusted for vehicle occupancy levels) produced dollar estimates.

We estimated that the proposed transit-land use plan would yield a net quantifiable benefit of around $28 million in the Year 2025 (in today's dollars). Quantifiable benefits were estimated to exceed annualized incremental capital and operating costs (relative to the "without" scenario) by 64 percent. The lion's share of monetizable economic benefits would accrue from travel time savings to motorists. Other, non-monetized benefits (e.g., regional economic growth; improved job accessibility for the poor; reduced public infrastructure outlays resulting from compact growth; etc.) would likely push the Year 2025 benefit-cost ratio above 2.0. We left it to elected officials to weigh the relative importance of these other factors.

Close

Coordinated transit and land use is which is widely embraced, but remains relatively little understood in a hands-on technical sense. Recent research provides benchmarks on what we might expect from integrated planning. However adapting and applying the information using traditional models in a case-specific context is no easy task. For the most part, planners need to be resourceful, drawing upon the best information available. Perhaps most helpful is establishing basic indicators of transit-land use sensitivities. Elasticities, such as the 0.197 value between station-area population densities and LRT boardings from TCRP H-1, are particularly helpful in conducting sensitivity and pivot-point tests. Triangulating forecasting approaches, as was done in Charlotte-Mecklenburg, also make a lot of sense. Clearly, if integrated transit and land-use planning and development is to gain acceptance and credibility, better analytical and forecasting tools will be needed. Some of the approaches adopted in Charlotte-Mecklenburg transit-land use study suggest promising pathways.

Robert Cervero, Ph.D, University of California, Berkeley