Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses (2022)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

44 Aging rail systems face a dilemma. These systems are often capacity-constrained all day, but par- ticularly in the peak hours. This gives such systems limited opportunities to conduct maintenance activities. How much do maintenance-related service cuts on weeknights and single-tracking impact ridership on the system? Should transit agencies weigh construction savings/efficiencies that can be achieved through single-tracking or shutdowns versus long-term ridership and revenue loss impacts? BART, or Bay Area Rapid Transit, is a heavy rail system operating in the San Francisco Bay Area, California. BART started operating in 1972 and gradually expanded the rail system. Today, BART serves 411,000 passenger trips per weekday at 48 stations on 112 miles of tracks, making it the fifth-largest rail operator in the United States. As shown in Figure 5-1, much of BART’s five main lines (Red, Yellow, Blue, Green, and Orange) operate on the same right-of-way. Four of these lines cross the Transbay Tube, thereby connecting the San Francisco Peninsula with the East Bay. The San Francisco Bay Area, where BART operates, has experienced profound transforma- tions in recent years. Between 2010 and 2018, the population of the nine-county region has grown by 8.5%, or 600,000 people (Green and Shuler, 2019). In the same time period, the number of jobs in the Bay Area grew by over 25% (Avalos, 2019). This economic growth has led to rising housing costs. According to the California Association of Realtors, housing affordability in the Bay Area dropped sharply following the Recession and has remained low ever since (Zaludova, 2018). The housing crisis has led to greater population growth in the East Bay counties of Almeda (10.4%) and Contra Costa (9.6%). This population and demographic shift has resulted in longer commutes to reach employment centers in San Francisco. The exclusive right-of-way on which BART trains operate has limited capacity. The Transbay Tube between San Francisco and Oakland already carries 27,000 passengers per hour during the peak, which is close to twice the volume of the Bay Bridge (Bay Area Rapid Transit, n.d.). How- ever, the Tube has already reached its maximum capacity of 23 trains per hour in each direction, leading to crowded conditions inside the vehicles and at stations. This bottleneck determines the frequency for the entire rail system. Since the five main BART lines share right-of-way, vehicles must be dispatched at a constant cadence. In other words, all lines must have the same frequency, even the Orange Line, which does not cross the bay. BART is planning to expand the capacity of the Transbay Tube by 30% as part of the Transbay Core Capacity Project, which is scheduled to be completed by 2028 at a total cost of $3.5 billion (Bay Area Rapid Transit, n.d.). The project consists of purchasing new vehicles, expanding car storage facilities, and replacing the existing fixed-block train control system with a new communications-based train control system. These improvements will allow BART to increase capacity by running longer trains more frequently. C H A P T E R 5 Examining the Peaking Phenomenon in Bay Area Rapid Transit Ridership

Examining the Peaking Phenomenon in Bay Area Rapid Transit Ridership 45   BART is studying plans to expand capacity across the bay with a second tunnel. According to a report by the Association of Bay Area Governments and the Metropolitan Transportation Commission, the new crossing would carry an additional 25,000 passengers at peak hour and cost between $30 billion and $50 billion. 5.1 Ridership Trends Figure 5-2 shows UPT and VRM relative to 2008. Both grew between 2002 and 2008, with a sharp fall during the Recession. Starting in 2010, both ridership and service levels started growing again. However, while UPTs reached pre-Recession levels in 2012, it wasn’t until late 2015 that the same happened for VRMs. Ridership then started declining in early 2016 and has been on a downward trajectory ever since. Meanwhile, service levels kept increasing until early 2019, when single-tracking had to be implemented to address urgent maintenance needs. Source: BART. Figure 5-1. Map of the BART system.

46 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses The decline in BART ridership is enigmatic because it coincides with growth in both jobs and population. To better understand how ridership has evolved by time period, Figure 5-3 shows the average hourly passenger trips by time of day for weekdays, Saturdays, and Sundays. Passenger trips presented in Figure 5-3 are linked, meaning that trips involving connections are counted only once, unlike UPTs. The trip time is recorded when passengers exit the station at their destina- tion. Each graph shows years from 2011 to 2019, with more recent years drawn in a darker line. Only 2019 is shown in red for easy identification. Figure 5-3 shows clearly the peaking phenomenon. On weekdays, 2019 has the highest rider- ship during peak hours and the lowest ridership during the off-peak. This effect corresponds to a long-term trend, with darker, more recent years clustered toward the top in peak hours and toward the bottom in midday and evening. On Saturdays and Sundays, 2019 has the lowest ridership at all times of day. This, too, corresponds to a trend over the last decade. Therefore, not only has ridership declined in recent years but also it has changed shape—growing in peak hours and decreasing at all other times. While it is clear that the recent ridership decline was driven by the off-peaks and weekends, the reasons for this shift are not entirely explained. Numerous external factors can account for this peaking phenomenon. The greater residential growth in the East Bay has generated more demand during commuting hours, while the rise of ride-hailing has provided an alternative to transit, which is particularly used at night and on weekends. There is also an internal factor that may have contributed to the trend: a reduction in nighttime and weekend service for track maintenance. The following section examines the rise of single-tracking and evaluates its impact on ridership. 5.2 Single-Tracking After close to 50 years of operation, the BART system—and particularly the electrical system— requires maintenance. However, since there are not enough hours to perform track work between midnight and 6 a.m., the transit agency must single-track. Single-tracking consists of closing the tracks for repair in one direction while leaving the other directional track open. Trains must wait 80 85 90 95 100 105 110 115 120 125 130 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 VR M a nd U PT R el ati ve to 2 00 8 Vehicle Revenue Miles Unlinked Passenger Trips Figure 5-2. Six-month rolling average of BART VRM and UPT.

Examining the Peaking Phenomenon in Bay Area Rapid Transit Ridership 47   Source (data): BART. Figure 5-3. BART ridership by time of day on weekdays, Saturdays, and Sundays by year. their turn to cross the track, which slows down operations. Any delay tends to propagate much faster during single-tracking because trains in both directions need to wait for the late vehicle. Single-tracking typically happens during the weekends and in the evenings, when ridership is at its lowest point. In order to enable single-tracking, BART reduced service on weekday evenings in February 2019. According to BART staff, and as shown in Figure 5-2, these were the first service cuts in several years. Figure 5-4 compares the service frequency of each line before and after February 2019.

48 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses Observable changes in service levels over time do not explain the peaking phenomenon. As shown in Figure 5-4, weekday frequency before 6 p.m. remained the same except for the Purple Line, which opened in February 2019 and only has two stations. In the evenings, however, service was cut on every route. The Yellow Line lost three train trips, while the Blue and Orange Lines lost two trips over the six-hour period. Because the routes only had two or three train trips per hour to start with, even losing two or three train trips over a six-hour period had an appreciable impact on headways. Although BART changed the weeknight schedule to accommodate single-tracking, not all week- nights had single-tracking. The service cuts on weekday evenings to accommodate single-tracking were in place between February 2019 and March 2020, when service was changed in response to Figure 5-4. BART service frequency by time of day for each line.

Examining the Peaking Phenomenon in Bay Area Rapid Transit Ridership 49   the COVID-19 pandemic. Therefore, the schedule information presented to passengers through the GTFS and related third-party applications distinguished between single-tracking days and regular days. Similarly, although BART implemented single-tracking on and off on weekends, published schedules showed service as if single-tracking was always in place. Figure 5-5 shows daily ridership over time in 2019. Weekdays are represented as squares, Satur- days as circles, and Sundays as triangles. Data on single-tracking were obtained from the events and track work records. Full single-tracking days are shown in blue, partial in green, and days with no single-tracking are shown in red. Full days are when single-tracking is applied between 8 a.m. and 6 p.m. Partial days are when single-tracking is only implemented for several hours; this typically happened in the evening. Figure 5-5 shows that single-tracking Saturdays and Sundays have lower ridership than regular weekends. Single-tracking weekdays, on the other hand, are indistinguishable from regular week- days. At the bottom of the graph, red circles and triangles protuberate over their blue and green counterparts. Interestingly, single-tracking weekend days seem to be uniformly distributed through- out the year, unlike single-tracking weekdays, which are concentrated in the summer. Figure 5-5 also shows that there is no discernible ridership difference on days with and without single-tracking. This could be because single-tracking undermines a seasonality, which would have otherwise abated summer ridership. However, the following explanation of endogeneity is more credible. BART planners decided whether to carry out single-tracking on weekends based on the antici- pated demand that would be affected. For example, the transit agency avoided single-tracking on days when concerts, parades, sporting events, or political rallies were expected to draw large crowds. Therefore, the relationship between ridership and single-tracking is endogenous. While single-tracking causes ridership to diminish, low anticipated ridership causes service planners to schedule single-tracking. Therefore, the high ridership on non-single-tracking days may be both a cause and a consequence of single-tracking. Source (data): BART. Figure 5-5. BART single-tracking days in 2019.

50 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses Another perspective is that single-tracking affects ridership in the longer term. Because single- tracking does not involve changing the schedule and would thus not be reflected in GTFS- based information sources, many passengers typically do not know whether their train will be single-tracked. Single-tracking not only makes trips longer but also makes the ridership experience less reliable. This is particularly true on trips that involve a transfer, in which case the penalty for single-tracking is doubled. Passengers, not knowing whether the rail system will be single-tracking the next time they need to travel, have to budget enough time to avoid arriving late at their destination in case single-tracking is unexpectedly happening. This additional cost imposed on passengers mitigates the value of public transit and could lead to long-term shifts toward other modes. 5.3 Conclusion While the service cuts on weeknights and single-tracking coincide with the ridership decline at nights and on weekends over the last year, they do not entirely explain the ridership trends. The analysis reveals that growth in the peaking phenomenon has been happening over the last decade, while BART ridership started to decline in 2016. What the service cuts and single- tracking demonstrate, however, is the capacity constraints imposed on a heavy rail system built in the early 1970s that requires maintenance. BART has limited opportunities to increase night and weekend ridership due to constraints on right-of-way and hours of operation. This balance between maintenance and service shutdowns is an issue that many transit agen- cies face, especially those with aging rail systems similar to BART, such as those in Boston, Philadelphia, and Washington, D.C. Agencies should carefully quantify and weigh the customer experience, equity, and ridership impacts of maintenance shutdowns against the maintenance cost savings afforded by the shutdowns. This type of analysis should be used to determine which maintenance activities warrant shutdowns and which should be restricted to non-service hours. More than simply a consequence of declining ridership—which transit systems across the country have also been experiencing—the peaking phenomenon is an issue in itself. Due to the limited capacity of the Transbay Tube at peak hours, it is imperative to spread the demand temporally across times of day and days of the week and spatially within the East Bay and San Francisco Peninsula. To spread out the demand temporally, the region has different potential levers available, including service design, enhanced information, and TDM. While it may be difficult for maintenance reasons Decades-old rail systems have capacity constraints that make maintenance problematic. While it is difficult to measure the impact of single- tracking on ridership, it should be considered when weighing maintenance timing. Endogeneity Identifying causal relationships is key to understanding the world around us and informing important decisions. A powerful tool available to data scientists is the examination of correlations in real-world data. Correlation is useful because it allows scientists to predict one variable based on another. However, correlation does not imply causation. Sometimes, an explanatory variable itself can be determined by the response variable. This happens in cases of reverse causation (when the response variable has a direct causal relationship with the explanatory variable) or simultaneity (when both variables are causally determined by some other, unobserved, variable). In those cases, scientists say the relationship between the explanatory and the response variables is endogenous.

Examining the Peaking Phenomenon in Bay Area Rapid Transit Ridership 51   to scale back single-tracking that reduces service levels, high-capacity bus service could provide regional mobility at night and on the weekend to supplement heavy rail service. Enhanced, high-quality passenger information is also an important tool to help passengers plan their trips, especially when service is unpredictable. A significant TDM tool is peak/off-peak pricing, which is used by some systems (e.g., WMATA) to create economic incentives to avoid the peak period; however, this tool is recognized as institutionally challenging. Alternatively, BART could work in partnership with local and regional governments to incentivize off-peak travel through various types of fare discounts. Employers could also be engaged to play a TDM role to help reduce the peaking phenomenon by allowing their employees to work flexible hours or work from home. The COVID-19 pandemic has impacted every aspect of urban transportation systems over- night. In order to minimize the risk of transmission, employers have encouraged workers to telecommute. Meanwhile, transit agencies have shifted service to minimize passenger crowding. While the urgency of these shifts was motivated by a global crisis, the experience gained in the process may reach beyond the moment. Passengers, employers, and transit agencies have learned that managing travel demand and supply is possible.