How do you predict how many will ride a train?

Deciding whether or not to introduce a new passenger train service or make changes to existing service comes down to one question: how many (more) people will ride the trains after the changes are made? This is true for both government transportation agencies and private companies. Ridership forecasting is a famously inexact science, and the entity performing the forecasting is liable to either overestimate or underestimate demand based on which approach best suits its interests. However, some methods are better than others, and new (and hopefully improved) methods are constantly being experimented with. Experts from state agencies, the Federal Railroad Administration (FRA), Amtrak and a consulting firm hired by Amtrak shed some light on the methods they use at a lecture session during the Transportation Research Board of the National Academies’ Annual Meeting in Washington, DC on January 12.

The chief difficulty in forecasting demand for intercity passenger trains, as opposed to commuter trains, is that non-work-related trips, which dominate intercity rail, are much less predictable than work-related trips. To further complicate matters, explained FRA Regional Planner Kyle Gradinger, travel trends are changing nationwide. Intercity buses are now a major competitor to Amtrak, which they weren’t a decade ago. Other noteworthy trends are declining vehicle miles traveled, and technology that is allowing people to more easily find travel information and be more productive when not behind the wheel.

There is also great variability in the ways people travel the “first and last mile” between the train station and their origins and destinations, particularly between the transit-rich Northeast and the rest of the country, where the automobile is the dominant “feeder” mode. A related complication comes from the fact that models assume that each station has ample parking, an assumption that — if realized — tends to frustrate efforts to build walkable transit-oriented development around stations.

To account for these difficulties, FRA is taking a step back and modeling overall travel demand within regions, looking at the network effects of ridership to create effective multimodal networks, not just corridors. At the corridor level, FRA seeks to develop a detailed ridership picture that would lead to dynamic service models, aided by data that states are already collecting for other purposes.

Gradinger noted that existing travel demand models, each developed and used by a different consulting firm, lack a uniform methodology. Few mature passenger rail services have been developed using solely one model. Yet FRA, state agencies and Amtrak rely on consultants because of their resource constraints, and because they are good at what they do, in spite of each having its own theories and approaches. Gradinger touted two specific FRA-led projects that will improve demand forecasting: “CONNECT,” a regional-level sketch planing tool for rail networks serving city pairs less than 800 miles apart that provides high-level ridership forecasts and cost estimates, and the NEC Future comprehensive planning effort. The latter incorporates a new interregional model that combines a new household travel survey with data from its own first-ever intercity bus survey of the nine Northeast states, and brings together the models used by transit agencies, state DOTs and Metropolitan Planning Organizations.

Amtrak is seeking ways to refine its analysis, which focuses on catchment areas within 30 miles of each station, said the company’s Vice President of Market Research and Analysis, Bill Sheridan. There are many parts of the country, particularly the interior West, where people may drive over 100 miles to reach an Amtrak station. Therefore, the company seeks to develop a different “catchment area statistic” for each station. Amtrak is also fine-tuning its analysis of the effects of a host of service changes on ridership, including on-time performance, station improvements and changes in on-board amenities. The railroad is also looking for better ways to account for the fact that demand profiles vary by station location as well as trip purpose, with many Amtrak trips not fitting easily within the “business or leisure” duality. 

Capacity is beginning to constrain ridership potential on most routes, Sheridan admitted. He added that, during the summer, the peak portions of long-distance train routes have higher average load factors than Northeast Corridor trains because the former serve multiple markets. 

The Northern New England Passenger Rail Authority (NNEPRA), the Maine state agency that supports and oversees the Amtrak Downeaster, compares each ridership data point — broken down by train number — to the same date the previous year, with data going back to 2005. This, combined with extensive first-hand observation by NNEPRA staff who ride the trains regularly, is used to create demographic profiles of passengers. The agency also keeps track of all special events happening along the line, particularly sporting events and other major happenings in Boston, which is the origin or destination of 86% of all Downeaster riders’ trips. Holidays and weather are also factored in. The fact that the train numbers of the weekend trains are different from the weekday trains helps, explains NNEPRA Executive Director Patricia Quinn.

Based on NNEPRA’s own surveys of riders, the authority decided to move the first weekday morning departure from Portland to 5:20 AM, instead of 6:00, to allow passengers to reach Boston by 8:00 AM. Ridership on that train jumped as a result — contrary to the agency’s original expectation, the surveys revealed that many riders are willing to endure a 4:00 AM wake-up call in order to be at work at 8:00.

NNEPRA also keeps an eye on the Downeaster’s competition, particularly the intercity buses with which the trains share the station in Portland, Maine. One bus company runs hourly service between Portland and Boston that is $10 cheaper than the train and takes 35 minutes less, if the traffic is good.

“Ridership forecasting is not a spectator sport,” Quinn explained. “You don’t do it once and call it done. You manage expectations throughout the year. Data collection is very important.” Combining the ridership data Amtrak collects with NNEPRA’s own passenger surveys allows the agency to draw “little maps to see where people come from to get to different stations.” “Each of our trains has a personality,” she continued. “We know all the train numbers by heart.”

The impacts of Section 209 of the 2008 federal passenger train authorization law have caused the Michigan Department of Transportation to focus more on revenue and increasing the farebox recovery ratio, rather than simply on getting more passengers to ride, explains Office of Rail Director Tim Hoeffner. He recalled the failure of the mid-1990s service reduction on the Chicago-Grand Rapids Pere Marquette to five times weekly. “We’ve proven that you can’t cut your way to prosperity,” he said.

Hoeffner claimed that there is a huge unmet potential for the Chicago-Detroit-Pontiac Wolverines. Actual ridership growth on that corridor over the past five years has been very close to his Office’s projections, but was below what was expected for the past two years, as work related to the Indiana Gateway Project on the Chicago-Porter section of the route has delayed trains and lengthened travel times. He said MDOT is working on contracting out demand forecasting to a private firm. “It’s very expensive and complex, but the bottom line is that the schedules that get published … are never the same as what the forecasts are built on. Engineers don’t get the break that economists and analysts get.”

Masroor Hasan, with the consulting firm Steer Davies Gleave North America, added that models should take into account what kind of information the traveler has available when evaluating his or her mode choice. He also emphasized that if public-private partnerships are to be increasingly relied upon to finance capital investments, then forecasts have to be robust enough to make investors confident that the project will generate a sufficient return on their investment. Therefore, Hasan suggested that forecasts be independently peer reviewed by panels incorporating all stakeholder views. 

Hasan pointed to a few recent examples of investment-grade or similar passenger rail studies: the Florida DOT’s Orlando to Tampa proposed high-speed rail corridor study, the Minnesota DOT’s Tier 2 Northern Lights Express (Twin Cities-Duluth) study, the Illinois DOT’s Chicago-St. Louis Tier-2 corridor study and the XpressWest (the proposed private Victorville-Las Vegas service) study. He also said his firm is working with Amtrak to better model train-level ridership and revenue incorporating train-level capacity constraints and the impact of on-time performance on ridership and revenue.