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.

9   Possible Causes of Ridership Decline Identified in the Literature Any analysis of factors affecting transit ridership should begin with a review of the previous literature. In order to isolate the different factors that affect ridership, this review is segmented into internal factors, which are controlled by transit agencies, and external factors. Both inter- nal and external factors are then further divided into traditional factors and emerging trends, creating four areas of possible ridership change that have been identified in the literature. An overview of the factors in each of these four areas is provided in Table 2-1. Although some broad themes are presented in the remaining sections of this chapter, a literature review by the research team is also available in TCRP Research Report 209: Analysis of Recent Public Transit Ridership Trends, which is the predecessor to this project. The existing literature identifies the important factors and the likely direction of each, but it is clear that a mix of factors is contributing to recent transit ridership trends in the United States, pushing ridership in competing directions. Many of the factors listed in Table 2-1 are described in more detail in the following sections. 2.1 Internal Traditional Factors The three primary areas under a transit agency’s control that have traditionally impacted rider- ship are service quantity, fares, and service reliability. 2.1.1 Service Quantity Service levels are the most important factor in ridership under control of the transit agency. There is a consensus in the literature that transit service levels—as measured in vehicle revenue hours (VRH) or vehicle revenue miles (VRM)—are the primordial factor affecting transit rider- ship (Dill et al., 2013; Kyte et al., 1988; Liu, 1993; Gomez-Ibanez, 1996; Kohn, 2000; Evans et al., 2004). VRH have been found to explain up to 95% of the variation in transit ridership (Taylor et al., 2009). More recently, a study by Boisjoly et al. (2018) identified vehicle revenue kilometers as the primary determinant of ridership in a panel regression study of 25 transit agencies from 2002 to 2015. By 2018, bus VRM had still not recovered their pre-2009 levels following the Great Recession service cuts. The relationship between ridership and service levels is not purely causal, however, as transit planners strive to plan service where they believe demand exists. TCRP Research Report 209 found that although there is a clear relationship between VRH and unlinked passenger-trips (UPTs) in 2012, the relative change between the two variables between 2012 and 2016 was loosely correlated at the metropolitan area level (Watkins et al., 2019). At the route-segment level between 2012 and 2018, Berrebi et al. (2019) found that ridership is inelastic to frequency (i.e., a 1% increase in frequency generates less than a 1% increase in ridership). The elasticity is lowest on the most frequent routes. C H A P T E R 2

10 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses 2.1.2 Fares Overall, increases in fares will modestly decrease transit ridership. Although sensitivity to fares can vary widely within the customer base, modest changes in fares have been found to affect ridership greatly (Liu, 1993; Kohn, 2000). Kain and Liu (1999) evaluated the factors that contributed to increasing ridership in Houston, Texas, and San Diego, California, in the late 1990s while transit ridership was declining across the United States. The authors concluded that increases in service, reduction in fares, and growth in employment and population contributed the most to increasing ridership. In a time-series regression analysis of seven transit agencies, Wang and Skinner (1984) found that the elasticity of ridership fares ranges widely, from −0.042 to −0.62. Using a similar methodology, Chen and Chen (2011) found fare elasticities of −0.4 in the short term and −0.8 in the long term. In a cross-sectional study of transit agencies through- out the United States, Taylor et al. (2009) estimated that fare elasticity was −0.42. In a study of both bus and rail ridership between 1990 and 2017 in Vancouver, British Columbia, Mahmoud and Pickup (2019) found that ridership elasticity to fare is −0.3. Therefore, while studies vary in their estimates and while several studies point to differences between modes and time frames, ridership is generally considered to be inelastic to fares. In other words, a 1% increase in fares typically generates lower than a 1% decrease in ridership. 2.1.3 Speed and Reliability Improved service reliability, including on-time performance, will increase transit ridership. Service reliability is a leading concern for both transit-dependent and choice riders (Krizek and El-Geneidy, 2007). Unlike other internal factors, which can be measured directly, the impact of reliability on ridership is driven by the experiences and perceptions of passengers. A clear measure of service reliability is schedule adherence. In a cross-sectional analysis of ridership at the route level in Los Angeles, California, Chakrabarti and Giuliano (2015) find that on-time performance is significantly correlated with bus ridership. In a similar study exploring ridership Internal External T ra di ti on al • Service quantity • Fares • Speed and reliability • Service concentration • Access to transit • Security • Service quality • Density • Population • Employment • Income • Gas prices • Commuting policies • Car ownership • Demographics E m er gi ng • Restructuring transit networks • Demand-responsive services, flex route services, and microtransit pilots and partnerships • New fare media and fare integration • Real-time information • Maintenance issues • Dedicated transit right-of-way • School and employer partnerships • Fare discounts or elimination • Gentrification • Aging population • Millennials • Telecommuters • Delivery services • Congestion and parking pricing • Shared mobility (ride-hailing, bike- sharing, car-sharing, scooters) Table 2-1. Factors affecting transit ridership.

Possible Causes of Ridership Decline Identified in the Literature 11   trends over time, researchers from the Massachusetts Bay Transportation Authority (MBTA) came to the same conclusion (Thistle and Zimmer, 2019). 2.2 Internal Emerging Factors Three emerging areas under the control of transit agencies that can influence transit ridership are network design changes, technology, and demand-responsive service. 2.2.1 Bus Network Redesigns Bus network redesigns increase ridership, but largely through increases in service and decreases in coverage. Several recent service-related efforts to increase transit ridership have consisted in restructuring bus networks to prioritize service concentration with higher frequency along specific corridors over geographic coverage (Houston; Omaha, Nebraska; Austin, Texas; and Columbus, Ohio). Called the “hottest trend in transit” by Governing magazine at the end of 2017, bus network restructuring is being considered by transit agencies across the nation. In 2020, LA Metro, Dallas Area Rapid Transit, Southeastern Pennsylvania Transportation Authority, and Washington Metropolitan Area Transit Authority (WMATA) were planning similar bus network redesigns (Hymon, 2017; Laughlin, 2017; Powers, 2017). In November 2017, Streetsblog USA wrote that “transit ridership is falling everywhere—but not in cities that redesigned their bus networks” (Schmitt, 2017). However, many of these bus network redesigns were accompanied by net increases in bus operating budgets, which may partly explain the ridership stabilization or increases (Byala et al., 2019). The network redesigns have also posed equity questions, as some have reported that low-income communities lost access while higher- income communities gained it (Flynn, 2015). 2.2.2 Technology Passenger information can increase ridership, while the impact of mobile ticketing is still unknown. Technology changes can improve transit ridership as well. The provision of real- time information in Chicago, Illinois, was found to correlate with an increase in ridership when controlling for service levels, employment, and gas prices (Tang and Thakuriah, 2012). A study by Brakewood, Macfarlane, and Watkins (2015) examined bus ridership changes in New York City in response to the gradual availability of real-time bus information. The study revealed a median ridership increase of 2.3%, with higher increases on the largest routes. Additionally, many transit agencies are investing in new fare payment systems to improve payment conve- nience and the rider experience, particularly for tech-savvy riders who are already using their smartphones to pay for many other goods and services. While mobile ticketing and other new payment technologies are likely to have positive ridership impacts, there has been limited prior research specifically studying the effects of new payment systems. 2.2.3 Demand-Responsive Services The impact of demand-responsive services on ridership is unknown. To provide greater transit access in low-density neighborhoods or in times of lower demand, there is increased interest in using demand-responsive transit services as an alternative to fixed route transit. Research has shown that in low-density areas, demand-responsive transit can service short trips faster (Qiu et al., 2015) and at a lower cost than fixed routes (Edwards and Watkins, 2013). Several transit agencies have implemented demand-responsive services either to reach the first mile/

12 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses last mile or to connect origins and destinations directly (Becker et al., 2013; Westervelt et al., 2018; Bliss, 2017a). However, quantitative research on the ridership implications of these programs is still lacking. 2.3 External Traditional Factors The previous section considered factors that are internal, that is, under the control of a tran- sit agency. The following sections consider factors that are external to the transit agency’s immediate control: economic factors, gas prices, demographic trends, and employer-based com- muting policies. It should be noted that although transit agencies can promote employer-based commuting policies—and TDM more generally—they do not control actual decisions made by employers. 2.3.1 Economic Factors Rising employment levels have a positive impact on transit ridership overall. The level of employment has a mixed effect on transit ridership. While greater employment generates more trips from commuting and consuming, it also leads to private vehicle purchase (Hendrickson, 1986; Liu, 1993). The overall effect of the employment rate, however, has been found to be positive in several studies (Gomez-Ibanez, 1996; Taylor et al., 2009). TCRP Research Results Digest 29: Continuing Examination of Successful Transit Ridership Initiatives identifies the rise of employment rates as the leading cause for nationwide ridership increases between 1994 and 1996 (Stanley, 1998). 2.3.2 Gas Prices Small changes in gas prices only moderately affect transit ridership. High gas prices, such as those in Europe, can have a much larger impact on encouraging transit ridership, but these prices are not found in the United States. There is limited evidence in the literature that gas prices substantially impact transit ridership (Sale, 1976; Dueker, 1998), though the value of the cross-elasticity of gas price to transit ridership has been found to vary substantially based on gas prices, urban form, mode, and time frame. In a time-series regression analysis of seven transit agencies, Wang and Skinner (1984) found that the cross-elasticity ranges widely, from 0.08 to 0.80. Maley and Weinberger (2009) found a cross-elasticity of 0.15–0.23 for city transit services and 0.27–0.38 for regional rail services. Yanmaz-Tuzel and Ozbay (2010) found a cross-elasticity of 0.12–0.22 in the short term, but this value drops to 0.03–0.18 in the long term. In a study of 11 Washington State counties over four years, Stover and Bae (2011) found a cross-elasticity of 0.17. Lee and Lee (2013) found that the cross-elasticity of ridership to frequency is 0.04 in sprawling metropolitan areas and 0.1 in compact regions. Nowak and Savage (2013) show that, as gas prices pass $3 or even $4, cross-elasticities increase to 0.28–0.38. Mahmoud and Pickup (2019) found a cross-elasticity of 0.08 in Vancouver. The general conclusion is that gas prices have relatively little impact on mode shift behavior, though they may cause some change in behavior in the short term when gas prices spike. A possible explanation for these generally low cross-elasticities is that persons with access to autos are relatively inelastic in their behavior, and gas prices are typically too low to motivate long-term vehicle purchases, which ultimately drives mode choice. 2.3.3 Demographic Trends Traditionally, younger people use transit more. Millennials have potentially conflicting impacts on transit usage. Demographic trends may play a significant role in shaping the pool of potential riders who may use transit (Coogan et al., 2018). Traditionally, transit ridership differs

Possible Causes of Ridership Decline Identified in the Literature 13   by age, with younger members of the population using transit more; this is often related to their stage in family lifecycle and income. Following retirement, daily trip-making for commuting purposes drops dramatically, which would affect transit in particular since it relies on commuters for a significant proportion of its ridership. The current aging population is such that members of the huge baby boom generation bubble have already started reaching retirement age in great numbers (Driscoll et al., 2018). In addition, demographic trends may be changing. There is fre- quent discussion of differences in travel behavior among millennials (born 1980–2000). On the one hand, they tend to be less auto-oriented in their preferences and to exhibit a propensity to use shared-use modes (Grimsrud and El-Geneidy, 2013; Grimsrud and El-Geneidy, 2014). At the same time, they are often avid users of modes that can be in competition with transit, such as ride-sourcing (Alemi et al., 2018). Issues of housing affordability may already be encouraging a move to auto-oriented suburbs as they settle into family households. 2.3.4 Employer Commuting Policies (Transportation Demand Management) Commuting policies and TDM programs encourage alternative modes, including transit. They can be successful in increasing ridership. Commuting policies, such as those included in TDM programs, are designed to create employer incentives that discourage the use of single- occupant vehicles for commuting by employees. Workplace policies have also been shown to alter employees’ commuting habits in more general ways. A 2017 study by Bueno et al. used a multinomial logit model to show that parking and driver mileage benefits correlated with decreased transit use, while transit benefits and employer-discounted passes correlated with higher transit use. This study was limited to New York and New Jersey, states with historically high transit use per capita. Similar research was conducted by Dong et al. (2016) in Portland, Oregon, and Block-Schachter and Attanucci (2008) in Boston, Massachusetts, both with similar results. There is limited research on transit benefit programs in small urban areas with a lower transit mode share. 2.4 External Emerging Factors As discussed repeatedly in the media and in the motivation for this TCRP study, there are many emerging external influences possibly impacting transit ridership. This section discusses six of the most significant factors: telecommuting and online shopping, gentrification, ride- hailing/ride-sourcing, car-sharing, bike-sharing, and dockless shared scooters. 2.4.1 Part-Time Employment, Telecommuting, and Online Shopping Non-traditional commuters may have less incentive to use transit, while decreasing non- commuting trips may result from changing shopping patterns. Telework, part-time and flex work schedules, and online shopping are becoming more prevalent and impacting the demand for travel and the times when trips are undertaken. Employment trends have been steadily decreasing the number of workers with traditional five-day, nine-to-five jobs while increasing the number of part-time workers and the number of workers who telework. According to a Gallup poll, 43% of Americans reported working remotely at least sometimes, a 4% increase since 2012 (Gallup, 2017). Telecommuters also reported working remotely more often; 75% reported working from home more than once per week, up from 66% in 2012. These changes naturally decrease the number of commuter riders on transit. This trend may also affect com- muters’ decisions to purchase monthly passes in favor of more flexible options, thus reducing non-commuting trips (Habib, 2017). In addition, the increasing growth in online shopping not only has resulted in the demise of major retail chains and shopping malls, but also is likely to

14 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses reduce the number of transit-based shopping trips. Delivery services such as Amazon and GrubHub have made shopping- and dining-delivery possible (Suel and Polak, 2018). Despite these trends, vehicle miles traveled are now at their highest point in history (Davis, 2017). 2.4.2 Gentrification Gentrification in cities generally decreases transit ridership. A potential contributing factor to the decreasing transit ridership is the economic displacement of low-income earners from dense urban centers to the suburbs. Despite some trends that were temporarily going in the opposite direction, suburbs have outpaced urban cores in growth rate (Frey, 2018). While cities are becoming denser, they are also becoming less affordable, which has led to the economic dis- placement of low-income and minority residents who constitute the primary transit user base (Florida, 2017). A study from Berrebi and Watkins (2020) found that a drop in the proportion of minority residents explains part of the ridership decline in Miami, Florida, but not in Portland, Oregon; Minneapolis, Minnesota; and Atlanta, Georgia, over the short time frame analyzed. 2.4.3 Ride-Hailing/Ride-Sourcing/Transportation Network Companies Early studies of ride-hailing impacts are mixed, therefore more research is needed. Since ride-hailing companies have started operating in U.S. cities, they have attained a 0.6% mode share in urban areas, which is substantial compared with the 1.7% and 1.1% mode shares of buses and passenger trains (Federal Highway Transit Administration, 2017). This trend, which coincides with the nationwide decline in ridership, has been investigated in multiple studies. In two surveys, respondents have reported that bus usage decreases while rail increases as a result of ride-hailing usage (Clewlow and Mishra, 2017) and that ride-hailing may replace or support transit differently for different trip purposes (Feigon and Murphy, 2016). Longitudinal studies conducted at the transit agency- or metropolitan area-level have come to diverging con- clusions. Several studies using data up to 2015 have found that the entry date of Uber had either a positive relationship with transit ridership or no statistically significant relationship (Hall et al., 2018; Boisjoly et al., 2018). Using a similar methodology but more recent data, Graehler, Mucci, and Erhardt found that ride-hailing was correlated with a decline in transit ridership (2019). Erhardt et al. found that between 2010 and 2016, congestion as measured by vehicle hours of delay increased by 60% in San Francisco, California, with two-thirds of that increase attributable to ride-hailing (2019). While the evidence thus far seems to point toward ride-hailing as a potential cause of nationwide ridership decline, this relationship is still not well understood. There remain many research questions surrounding the competition or complementary dynamics between transit and ride-hailing. 2.4.4 Car-Sharing Car-sharing has a complicated effect on ridership, possibly enabling a car-light lifestyle but sometimes substituting for transit use. The literature exploring the impacts of car-sharing on transit usage thus far has mixed results. Some studies reported that households that have access to car-sharing are using transit less than before (Martin and Shaheen, 2011; Sioui, Morency, and Trépanier, 2013). However, a report combining 15 studies found that car-sharing members’ transit usage increased between 13.5% and 54% after becoming members (Shaheen et al., 2009). A study of car-sharing in San Francisco showed that car-share members use transit for 14.5% of their trips compared to 10.3% for non-members (Clewlow, 2016). TCRP Report 108: Car-Sharing—Where and How It Succeeds reported that 20% of car-sharing trips were accessed by transit (Millard-Ball et al., 2005). Other studies reported that some car-sharing users substitute transit trips with car-sharing, while others are using transit more since they either reduced car

Possible Causes of Ridership Decline Identified in the Literature 15   ownership or used transit to access car-sharing. These mixed effects were noticed in Philadelphia and San Francisco (Cervero et al., 2007; Lane, 2005). 2.4.5 Bike-Sharing Bike-sharing can increase rail ridership at outlying stations, but it decreases bus rider- ship. Research in the literature has investigated the impact of bike-sharing on transit usage with surveys and empirical models. The main finding from surveys is that bike-sharing is both a competitor and complement to transit, as it replaces transit trips in dense areas and serves as a first-mile/last-mile connector in the suburbs (Buck et al., 2013; Fuller et al., 2013; Martin and Shaheen, 2014; Shaheen et al., 2013; Shaheen et al., 2014). Studies based on empirical models have quantified the impact of bike-sharing on transit ridership in New York City and Washington, District of Columbia. In New York City, it was found that every 1,000 bike-sharing docks along a bus route is associated with a 2.42% decrease in daily unlinked bus trips (Campbell and Brakewood, 2017). In Washington, an early study reported that a 10% increase of Capital Bikeshare rider- ship corresponds to a 2.8% increase in Metrorail ridership (Ma et al., 2015). Another study in Washington concluded that the impact of bike-sharing on Metrorail ridership was negative for stations located in core neighborhoods and positive for stations located in peripheral neighbor- hoods (Ma and Knaap, 2019). These findings indicate that bike-sharing both substitutes for and complements transit usage to varying degrees, depending on location. 2.4.6 Dockless Scooters The magnitude of the impact of dockless scooters on transit ridership is unknown. Dockless scooters made their entry to U.S. cities in summer and fall 2017. Due to the recency of this phenomenon, the research so far is based on surveys. Respondents to these surveys see shared scooters as a complement to public transit. Populus (2018) reported that in 11 major U.S. cities, 70% of the sample surveyed see electric scooters as a complement to public transit. The National Association of City Transportation Officials (NACTO) reported that in 2018, 25% of scooter trips are connections to transit (2019). Although scooters may also be competing with transit, their impact in filling the first-mile/last-mile gap has been reported to be greater in magnitude. In San Francisco; Denver, Colorado; Arlington, Virginia; and Bloomington, Indiana, 34%, 20%, 9%, and 4% of survey respondents, respectively, reported using scooters as a connecting mode to/from transit; meanwhile, only 15%, 7%, 5%, and 2%, respectively, reported substituting transit trips (San Francisco Municipal Transportation Agency, 2019; Denver Public Works, 2019; DeMeester et al., 2019; Baltimore City Department of Transportation, 2019). These results indi- cate that scooters may be enabling more ridership than they substitute. These findings, however, are only based on surveys and may be impacted by selection bias. 2.5 Conclusion Three internal traditional factors (those within a transit agency’s control) have traditionally explained changes in transit ridership, and all three still impact ridership levels. Service levels are the most important factor in ridership under control of the transit agency—increasing service traditionally has increased ridership. Improved service reliability, including on-time perfor- mance, will increase ridership. Overall, increases in fares will modestly decrease transit ridership. Similarly, there are four external factors that traditionally have impacted transit ridership. Rising employment levels have a positive impact on transit ridership overall. Small changes in gas prices only moderately affect transit ridership. Demographic trends, such as the aging and retirement of baby boomers, are decreasing the number of transit commuters, while the overall

16 Recent Decline in Public Transportation Ridership: Analysis, Causes, and Responses impact of millennials is unclear. TDM-related commuting policies to encourage transit usage and discourage auto usage can be successful at increasing ridership. As will be discussed in the analysis presented in Chapter 3, although these traditional fac- tors explain some of the recent ridership decline, they do not explain all of the changes being experienced by the industry. The impacts of emerging trends in technology, travel behavior, and transport policy are still not fully understood. Three new trends are emerging within the control of the transit agency that may impact transit ridership. The literature has shown that bus network redesigns increase ridership, largely through increases in service. Passenger information can increase ridership, while mobile ticketing is yet unknown. The impact of demand-responsive services on ridership is unknown. Finally, telecommuting, part-time employment, and online shopping will continue to decrease transit’s potential market, as will gentrification trends in many cities. The impacts of new modes— such as ride-hailing, bike-sharing, and car-sharing—are unclear. These trends may require cities to try new techniques and/or new partnerships. Traditional factors, such as service levels and fares, can explain some recent changes in transit ridership, but emerging trends must be better explored to understand the current trends.