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.
10 gentrification in urban cores has displaced transit-dependent populations to suburbs, and wealthier groups who are less likely to take transit have been taking their place. Although some suburbanites may use transit, their usage patterns will differ. ï· There are a growing number of resources that replace the need to make trips. Telecommuting and working from home are trends that have grown considerably in recent years, driving down the need for monthly transit passes. Delivery services such as Amazon or GrubHub make trips to stores and restaurants less necessary and frequent, and are particularly prevalent in urban areas well served by transit. ï· Shared mobility services are growing in popularity and likely have mixed effects on ridership. Bike and carsharing services make auto ownership less necessary, but there is evidence that they may be replacing transit trips. Some transit agencies and city officials are skeptical of integration with these services, as they see them as competitors. ï· There is evidence that Transportation Network Companies (TNCs) replace transit trips, particularly outside of peak hours. TNCs, like Uber and Lyft, are used for both recreational purposes and commuting, although mostly for off-peak and airport trips. However, many users report that these services replaced their transit trips. Overall, TNCs may add auto trips to the road, and raise vehicle miles traveled. ï· There is also evidence that Transportation Network Companies complement transit, particularly for rail systems. TNCs have the potential to serve as last-mile connections to rail and BRT systems and may help enable a transit lifestyle. Many cities have begun supplementing their demand-responsive service with TNC services to bridge system gaps. ï· Transit agencies have been upgrading technology in an attempt to win back riders. Improvements in real-time information has been shown to boost transit ridership slightly. Fare technology that improves simplicity and speeds up buses is being implemented in several cities, with limited results on the ridership effects of these changes. ï· Bus networks are being restructured to provide more concentrated service and attract riders. This trend consolidates low-frequency meandering services into high- frequency direct services, bringing more residents closer to high-frequency bus lines. Bus ridership effects have been slightly positive, but with limited results at this point. ï· Overall, there is little consensus as to the full picture describing recent transit ridership declines. There are a multitude of candidate factors, from competing services like TNCs to societal factors like gentrification. More research is needed to understand the impact of multiple factors, especially new trends in transportation, on transit ridership.
11 Chapter 2: Research Approach Transit agencies in the United States operate in a wide variety of environments, from small towns to mega regions, where decades of urban development have shaped the way people travel. This context affects not only the contributors to changing transit ridership, but also which strategies may be effective at offsetting ridership declines. While the overall ridership trend is pointing downward, it is important to identify sub-trends in order to grasp the full implications. Identifying the characteristics associated with transit ridership decline is also necessary to effectively target its root causes. Discerning the sub-trends is particularly relevant because the largest transit agencies account for a disproportionate share of ridership; the New York MTA alone contributed 33% of 2015 unlinked transit passenger trips in the U.S. The ridership decline could be attributed to few large transit agencies, for example due to extended rail closure; or it could be attributed to many small ones, for example due to urban migration; or it could be attributed to both. Furthermore, any analysis of averages would skew towards the largest regions and overlook ridership trends in smaller ones. Organizing transit agencies into groups of peers is therefore necessary to compare the evolution of transit ridership over time. Therefore, the research presented in this report was organized around two sets of clusters that group transit agencies according to similar operating environments and service characteristics. The clustering method is explained below. Using the clusters, national ridership trends were identified and graphed along with changes in population, transit vehicle revenue miles, and zero-vehicle households. Then, ten case study transit agencies were selected across the clusters to look at route- level ridership change within the transit agency. Clustering The first step of this analysis of ridership trends is to classify transit agencies with similar operating environments and service characteristics. A full description of the methodology used is described in a Transportation Research Record publication titled Comparing Transit Agency Peer Groups using Cluster Analysis by Ederer, et al. Transit regions were clustered into groups of peers on the basis of metropolitan area population, percent of population living in a dense area, percent of zero vehicle households, and transit operating expenses. Two cluster analyses were performed: one for transit services in mixed traffic, and one for services in a dedicated right-of-way. The mixed traffic and dedicated right-of-way mode categories were separated based on National Transit Database data. ï· Mixed traffic regions included all metro areas operating intra-city bus, commuter bus, bus rapid transit, and streetcar service.
12 ï· Dedicated right-of-way modes included heavy rail, light rail, monorail, and hybrid rail. Dedicated right-of-way services only included systems with 1 million or more unlinked passenger trips per year. Transit agencies that operate mixed and dedicated right-of-way service were included in both clusters. Metrics attributed to different modes were split according to mode for each clustering. This method captures the differences in operation and funding logistics that may be present for different modes within the same transit agency and region. With the understanding that many transit agencies operate in the same city, and that riders have little discretion for the specific transit agency operating a service, we found it useful to group regions rather than transit agencies. Transit providers within a region often compete for the same riders or connect groups of riders together, so pooling all of the transit service in a region provides a much more useful glimpse into particular ridership trends in a city than an agency-by-agency analysis. We clustered regions based on their core-based statistical area, often known as metropolitan or micropolitan statistical areas. This unit was chosen as it has the most data availability for any regional metric from the US Census. American Community Survey 5-year estimates were used for the years 2012 and 2016, as well as transit data from the National Transit Database supplemented with data from the American Public Transportation Association. The availability of timely data was a limitation of the study, as 2016 data was the most recent available at the time of analysis. Downward trends in transit ridership have continued into 2017 and 2018 with some cases being even more substantial than what is shown in this report. Clusters â Mixed Traffic Modes The resulting clusters are described below. Figure 4 shows a map of mixed traffic regions color-coded by cluster. In all cluster solutions, the New York City metropolitan area was an outlier. It was not included in this analysis. ï· Cluster 1 â Mid-sized, transit-oriented â Features older industrial cities that are typically in Northeast and Midwest that have declined in population in the past several decades. These areas have a relatively high number of zero vehicle households and are typically small to midsize metro areas. Example cities include Albany, Baltimore, Pittsburgh, and Cleveland. ï· Cluster 2 â Mid-sized, auto-oriented â Features primarily smaller, recently developed cities in the Midwest and South with low percentages of people living in zero vehicle households. Example cities include Indianapolis, Kansas City, Charlotte, and Nashville. ï· Cluster 3 â Sprawling small towns â Consists of the smallest cities operating fixed route transit service and includes a disproportionate number of âcollege towns.â The metro areas in this cluster are the least dense, least populated, and spend the least on transit of the transit
13 agencies included in this analysis. Example cities include Lansing, Burlington, Blacksburg, and Knoxville. ï· Cluster 4 â Sprawling metropolis â The cities in this cluster are sprawling, large cities that have a low percentage of zero vehicle households. Operating expenditures in this cluster reflect the large population of these areas. Example cities include Atlanta, Houston, Denver, and Phoenix. ï· Cluster 5 â Dense metropolis â Consists of the largest metro areas in the county. Metro areas in this cluster are very dense and spend substantially more on bus operations than regions in other clusters. Example cities include Boston, Philadelphia, Chicago, Seattle, and Miami. FigureÂ 4:Â MapÂ ofÂ MixedÂ TrafficÂ TransitÂ RegionsÂ byÂ ClusterÂ Clusters â Dedicated Right-Of-Way Modes The resulting dedicated right-of-way clusters are described below. Figure 5 below delineates the clusters for metropolitan areas operating dedicated right-of-way services with at least 1 million trips per year.
14 ï· Cluster A â Los Angeles â The Los Angeles metropolitan area is an outlier in this grouping. It is unusually large with a higher percentage of people in dense areas, but with very low investment in dedicated right-of-way service. ï· Cluster B â Dense metropolis â Includes Chicago, Boston, Philadelphia, San Francisco, and Washington D.C. These are large metro areas with extensive transit systems and large commuter rail networks. ï· Cluster C â Mid-sized, dense â Consists of cities that are relatively small, compact, and with a high number of zero vehicle households. This includes former industrial hubs in Baltimore, Buffalo, Cleveland, and Pittsburgh. ï· Cluster D â Mid-sized, dense, auto-oriented â Consists of medium sized metro areas that are mainly in the western areas of the country such as San Jose, Portland, Seattle, Phoenix, Sacramento, Denver, and San Diego as well as Miami. These cities have low percentages of zero vehicle households, but a high proportion of population living in dense census tracts. ï· Cluster E â Sprawling metropolis â Consists of sprawling large metro areas with relatively few dense census tracts, many of which are located in the southern (Atlanta, Dallas, Houston, Charlotte) and western (Salt Lake City, Minneapolis, St. Louis) regions of the U.S. Figure 5 presents the clusters in the form of a dendogram, in which the regions most closely related are shown as connected by a line. Cluster A (Los Angeles) is therefore more closely related to Cluster B (Dense Metropolis) than to the other clusters. Similarly, Clusters D (Mid-sized auto- oriented) and Cluster E (Sprawling Metropolis) are more closely related to each other than the other clusters, and so on.
15 FigureÂ 5:Â DendogramÂ ofÂ DedicatedÂ RightâofâWayÂ ClustersÂ Ridership Trends As the major factors traditionally influencing transit ridership, it is important to understand how ridership is changing according to changes in service levels, population, and transit-dependent population. Therefore, for each mixed traffic and dedicated right-of-way cluster, a trend analysis was performed to examine the relationship between transit ridership and these three factors. In all cases, transit ridership was defined by unlinked passenger trips (UPT). Service levels are represented by transit vehicle revenue miles (VRM), although multiple similar measures were tested. Population is represented by one-year American Community Survey (ACS) estimates. Transit-dependent population is represented by zero-vehicle households from the ACS. Additional factors were considered, but due to data limitations, these three were the most reliable across multiple regions. Appendix B clarifies the data limitations the study team faced in the analysis. With regard to transit vehicle revenue miles, other service level variables were considered, but all service level variables were very closely linked, leading the study team to conclude that only one was necessary for further analysis. Transit Agency Strategies and Case Study Selection There is little existing peer-reviewed research on strategies that transit agencies have taken to combat the declines in transit ridership. Therefore, news articles and transit agency reports were examined to get a picture of strategies being undertaken and the degree to which they have been successful. E A B C D
16 Taking into account both the transit ridership trends and factors influencing those trends and the strategies transit agencies are using to combat ridership change, ten transit agencies were selected to conduct case studies. Table 1 lists the ten transit agencies and their associated clusters for mixed traffic modes and dedicated right-of-way modes. Five of the transit agencies have both dedicated right-of-way and mixed traffic modes; all five mixed traffic mode clusters are represented; and all dedicated right-of-way clusters except Los Angeles are represented. TableÂ 4:Â CaseÂ StudyÂ TransitÂ agenciesÂ Transit Agency City Mixed Traffic Cluster Dedicated ROW Cluster Connect Transit Bloomington-Normal, IL 2 N/A IndyGo Indianapolis, IN 2 N/A Pinellas Suncoast Transit Authority St. Petersburg, FL 2 N/A Spokane Transit Authority Spokane, WA 2 N/A Greater Portland Transit District Portland, ME 3 N/A Maryland Transit Administration Baltimore, MD 1 C Metro Transit Minneapolis-St. Paul, MN 1 E Metropolitan Transit Authority of Harris County Houston, TX 4 E Massachusetts Bay Transportation Authority Boston, MA 5 B King County Metro Seattle, WA 5 D
17 Chapter 3: National Ridership Trends Based on the literature and industry knowledge, the major factors traditionally influencing transit ridership are changes in service levels, population, and transit-dependent population. For each mixed traffic and dedicated right-of-way cluster, a trend analysis was performed to examine the relationship between transit ridership and these three factors. In all cases, transit ridership was defined by unlinked passenger trips (UPT). Service levels are represented by transit vehicle revenue miles (VRM). Population is represented by one-year American Community Survey (ACS) estimates. Transit-dependent population is represented by zero-vehicle households from the ACS. Population, zero vehicle households, and transit service levels were plotted against transit ridership to determine whether they were similar in magnitude and direction. In each case, the relationship between transit ridership and its determinants is first evaluated using only 2012 data. This point-in-time analysis helps understand the steady-state effects each factor has on transit ridership after decades of interaction. Of course, the causal relationship goes both ways as transit ridership could directly or indirectly affect population size, share of zero-vehicle households, and transit service levels. In the second part of the analysis, the percentage change in transit ridership is compared to the percentage change in each explanatory factor between 2012 and 2016. This analysis helps understand their relationship in the short term. Exploring how factors change together within a four year period can provide important insights on the potential causal relationships that have been driving ridership down. In all of the figures, regions are abbreviated with three letter codes, shown in Appendix C. Ridership Trends Analysis for Mixed Traffic Modes Using the Clusters 1 through 5 as explained in the methodology section, trends in transit ridership versus population, zero-vehicle households, and service levels at a point in time as well as trends in change in transit ridership versus changes in these factors are described in the following section for the mixed traffic modes. For more information about which regions are in each cluster, see Appendix C.
18 Population â Mixed Traffic Modes While transit ridership is declining nationally, the population of urban areas overall is at its highest point in history. Although suburbs have been growing at a faster pace in recent years, urban cores have increased in population every year since 2006 (Frey 2018). These trends make the recent decline in transit ridership even more alarming because they indicate that ridership per capita has been falling at an even faster rate than ridership. One potential explanation for the transit ridership decline is that population growth has been concentrated in sprawling metropolitan areas, particularly in the Sun Belt, while denser cities, particularly in the Rust Belt, have lost population. According to a study by Driscoll, et al (2017), urban migration away from cities with strong transit markets may be a leading cause of the ridership decline. The relationship between population and transit ridership in each cluster is first evaluated in 2012 as shown in Figure . Clusters 1 and 5, which contain transit supportive regions, population explains a large part of the variation in ridership among regions. In Cluster 1, which contains mid-sized MSAs, only San FigureÂ 6:Â 2012Â TransitÂ RidershipÂ vsÂ PopulationÂ forÂ MixedÂ TrafficÂ ModesÂ
19 Juan, Detroit, MI, and Honolulu, HI, deviate from the linear relationship between population and ridership. In Cluster 5, which contains large MSAs, the relationship is also clear, although heavily influenced by Los Angeles and Chicago. Clusters 2, 3, and 4, which are all car-oriented MSAs have diverging trends. While Cluster 2, which contains mid-sized MSAs, shows a clear positive relationship between population and ridership, Cluster 4, which contains larger MSA, does not. Cluster 3 (small towns) seems to be showing a trend, but it is more difficult to identify because a large subset of the cluster is highly homogeneous. The 62% of regions that have less than a million in population and less than 5 million unlinked transit passenger trips seem to be uniformly distributed. However, the ridership of satellite regions orbiting around this core group tends to increase with population. Overall, the trend between population and transit ridership is quite strong in every cluster except for Cluster 4, sprawling metropolises. FigureÂ 6Â (continued):Â 2012Â TransitÂ RidershipÂ vsÂ PopulationÂ forÂ MixedÂ TrafficÂ ModesÂ
20 Population Change â Mixed Traffic Modes FigureÂ 7:Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ PopulationÂ forÂ MixedÂ TrafficÂ ModesÂ In order to understand how the change in ridership between 2012 and 2016 relates to population, Figure shows ridership change against population change in percentage. Despite the strong trends in Figure , the short term trends in Figure are mixed. In some cases, metro regions with dense urban cores lost the most ridership and gained the least population. Areas in Cluster 1 (mid-sized transit-oriented) lost the most population with 20 out of 47 metro regions experiencing a net population loss. Metro regions in Cluster 5 (dense metro) all grew by less than 7%. In both clusters, however, transit ridership change seems unaffected by the decline in population. Clusters 2 and 3 (mid- sized auto-oriented and sprawling small towns) have no clear trend either. Metro regions are scattered around a negative change in ridership that is not affected by changes in population. In both clusters, the trend line is flat, which indicates that the change in transit ridership is uncorrelated with change in population. Despite population growing by more than 4% for every MSA in Cluster 4 (sprawling metro), there is also no relationship.
21 FigureÂ 7Â (continued):Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ PopulationÂ forÂ MixedÂ TrafficÂ ModesÂ Overall, there is a strong relationship between transit ridership and population but almost none between ridership change and population change. In every cluster except for Cluster 4, population explains a large portion of the variation in ridership. It is interesting to note that Clusters 2, 3, and 4 have gained the most population perhaps at the expense of Clusters 1, and 5. However, if population change was a significant factor of transit ridership for mixed traffic modes, then these trends would be reflected within each cluster in Figure 5, particularly Clusters 1 and 5. We therefore conclude that although population is an indication of transit ridership, other factors than recent population changes are having an overwhelming effect on transit ridership change. It may be that change within a population, such as demographics, are having a larger impact than overall population change.
22 Zero Vehicle Households â Mixed Traffic Modes The proportion of households with zero vehicles is an important indicator of transit ridership because it reflects the medium to long term propensity of individuals to ride public transportation. Zero vehicle households can be delineated into two groups : ï· car-free households, where residents choose to live without a car, and ï· car-less households, where residents lack access to a vehicle for physical or financial reasons. A recent study based on 2012 California Household Travel Survey found that 79% of zero vehicle households are car-less (Brown 2017). In general, the total lack of cars in a household, is more closely associated with constraint than choice. These households, therefore constitute a population sometimes refered to as captive transit riders, because they have no other accessible means of transportation. Larger households can also be one-vehicle households, thus necessitating travel by other means for most members of the household, however readily available data to quantify these households in every region is more difficult to obtain. Therefore, zero- vehicle households is used as a surrogate for those without automobile access. FigureÂ 8:Â 2012Â TransitÂ RidershipÂ vsÂ 2012Â PercentÂ ofÂ ZeroÂ VehicleÂ HouseholdsÂ
23 The proportion of zero-car households should also be understood in the context of density. Density determines whether people can live without a car and still have access to alternative transportation options. While Clusters 1 and 2 have the same proportion of population living in transit supportive density, the proportion of households without cars is almost twice as great in Cluster 1. The same can be said about Cluster 4 and 5, which have similar densities but very different proportions of zero- car households. These numbers suggest that regions in clusters 1 and 5 have large proportions of transit- dependent households. Figure shows the percent of zero vehicle households against unlinked passsenger trips in each cluster. In Clusters 1, 2, and 3 (lower population MSAs), the regions with the greatest transit ridership have medium and in some cases low proportions of zero vehicle households. These regions include Honolulu, HI, and Baltimore, MD, in FigureÂ 8Â (continued):Â 2012Â TransitÂ RidershipÂ vsÂ 2012Â PercentÂ ofÂ ZeroÂ VehicleÂ HouseholdsÂ Cluster 1, Orlando, FL, and Austin, TX, in Cluster 2, and St. Louis, MO, and Durham, NC, in Cluster 3 which all have far greater levels of ridership than the average in their cluster, yet have relatively low zero-vehicle households compared to other cities. Besides these outliers, there is a slight positive trend in all three clusters. However, it is apparent that the proportion of zero vehicle households accounts for a small share of variation in transit ridership. There is also not a strong relationship between transit ridership and the proportion of zero vehicle households in Clusters 4 and 5 (higher population MSAs), although the more car-oriented large metropolitan areas in Cluster 4 have a slightly stronger relationship than the more transit-oriented large metropolitan areas in Cluster 5.
24 Change in Zero Vehicle Households â Mixed Traffic Modes FigureÂ 9:Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ ZeroÂ VehicleÂ HouseholdsÂ forÂ MixedÂ TrafficÂ ModesÂ While the proportion of zero vehicle households at a point in time reflects the steady-state of economic, land-use, and transportation forces, the increase in car ownership has been identified as a major cause of transit ridership decline. A study from the Southern California Association of Governments suggested that the decrease in zero-vehicle households was the primary cause of transit ridership decline in the Los Angeles greater area (Manville et al. 2018). To evaluate this trend at the national level, Figure shows the change in transit ridership against the absolute change in percentage of households with zero vehicles between 2012 and 2016. Unlike the 2012 analysis, there is a relationship between the change in transit ridership and the change in zero vehicle households in Cluster 4 (sprawling metros) and Cluster 5 (dense metros), but not in Clusters 1, 2, and 3 (lower population MSAs). The relationship between the change in transit ridership and zero vehicle households is rather flat in Clusters 1, 2, and 3, with regions spread widely and almost symetrically around.
25 FigureÂ 9Â (continued):Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ ZeroÂ VehicleÂ HouseholdsÂ forÂ MixedÂ TrafficÂ ModesÂ In the large metro areas, Clusters 4, and 5, however, the change in zero vehicle households is associated with change in transit ridership. In Cluster 4, only Las Vegas, NV, increased in proportion of zero-vehicle households, and in Cluster 5, Seattle, WA, was the only region not to decline in zero vehicle households except for Boston, MA, which did not substantially change. Overall, the decline in transit ridership is therefore connected with the decline in proportion of zero-vehicle households in large metro areas (Clusters 4 and 5) but not in smaller ones (Clusters 1 to 3).
26 Transit Service â Mixed Traffic Modes The amount of service provided is one of the few levers available for transit agencies to affect ridership. It is therefore important to evaluate the relationship between ridership and service levels both at a point in time and as a change over time. In order to better understand the base-case relationship between transit ridership and service levels, Figure shows 2012 transit ridership against 2012 transit service levels in each cluster. In every cluster, the relationship between transit ridership and transit service levels is both clear and strong. In clusters 1 and 2, transit service levels explain almost all of the variation in transit ridership. In both clusters, transit service levels span a wide spectrum with the largest regions having 50 times more transit service than the smallest ones. The relationship between transit ridership and transit service levels holds true at all levels of transit service levels. It is interesting to note that the trend line has slope equal to 3 in Cluster 1 and slope equal to 2 in Cluster 2. These results suggest that each marginal vehicle revenue mile is associated with more transit ridership in mid-sized transit-oriented Rust Belt regions than in car-oriented, and for the most part Sun Belt, regions of similar sizes and densities. FigureÂ 10:Â 2012Â TransitÂ RidershipÂ vsÂ 2012Â TransitÂ ServiceÂ LevelsÂ forÂ MixedÂ TrafficÂ ModesÂ
27 In cluster 3, the trend is also clear, but there is not as much spread in transit service levels as in Clusters 1 and 2. There are 9 regions with much more transit vehicle revenue miles and transit ridership than the rest of the cluster. These are the same regions that had overwhelming transit ridership for their proportion of zero vehicle households in the last analysis. All other regions within the cluster are compact between zero and three million transit vehicle revenue miles. The group of compact regions shows a relationship between transit ridership and transit service levels, which extends to outlying regions with far greater transit service levels, thereby confirming the trend. Although there is also a clear relationship between transit ridership and transit service levels in Clusters 4 and 5, it does not explain as much of the variation in transit ridership. This is especially true in Cluster 4 with Las Vegas, NV, and Dallas, TX, for which transit service levels do not explain transit ridership well. FigureÂ 10Â (continued):Â 2012Â TransitÂ RidershipÂ vsÂ 2012Â TransitÂ ServiceÂ LevelsÂ forÂ MixedÂ TrafficÂ ModesÂ Besides Boston, MA, all regions in Cluster 5 provided more transit service in 2012 than even the largest region in Cluster 4, which was Houston. It is also worth noting that the slope of the relationship between transit ridership and transit service levels is three times greater for Cluster 5 than for Cluster 4. These results indicate that transit service levels contribute far more ridership in the large transit-oriented metropolitan areas of Cluster 5 than in more car-oriented sun-belt regions of Cluster 4. Figure shows that there is a strong relationship between transit ridership and service levels at a point in time. These results suggest that transit service levels may be a strong influencer of transit ridership, but it is important to also evaluate the influence of changes in transit service levels in the next section to explain the ridership decline since 2012.
28 Change in Transit Service Levels â Mixed Traffic Modes FigureÂ 11:Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ TransitÂ ServiceÂ LevelsÂ forÂ MixedÂ TrafficÂ ModeÂ As shown in Figure 11, there is a definite relationship between change in transit ridership and change in transit service levels in Clusters 1 and 2, and to a lesser extent, Cluster 3. ï· In Cluster 1, the spread around the trend line is wide, but consistent. Every region where ridership has grown has increased transit service levels. ï· In Cluster 2, the relationship explains a large part of the variation in transit ridership. Regions with the greatest ridership growth have increased transit service levels the most and regions with the greatest ridership fall have reduced transit service levels the most. ï· In Cluster 3, several regions where both transit ridership and transit service levels increased dramatically between 2012 and 2016 drive the relationship. There are, however, outliers such as Chico, CA, Huntsville, AL, and Elizabethtown, KY, where ridership increased despite slight reductions in transit service levels and regions such as Fayetteville, MO, Baton Rouge, LA, and Port St. Lucie, FL, where vehicle revenue miles increased by more than 75% but transit ridership did not substantially increase.
29 FigureÂ 11Â (continued):Â ChangeÂ inÂ TransitÂ RidershipÂ vsÂ ChangeÂ inÂ TransitÂ ServiceÂ LevelsÂ forÂ MixedÂ TrafficÂ ModeÂ Besides the strength of the relationship, the intercepts are also interesting. The intercept of the trend lines is the amount of ridership change that should be expected if transit service levels had not changed. The intercept is -11% for Cluster 1, - 9% for Cluster 2, and -8% in Cluster 3. These results indicate that small to mid-sized regions that didnât change transit service levels between 2012 and 2016 should expect 8-10% loss in ridership. Although there is a definite relationship between the change in transit ridership and the change in transit service levels, there is also a systematic effect driving transit ridership down in transit agencies across clusters irrespective of service levels. In Clusters 4 and 5, there is no discernable relationship between the change in transit ridership and the change in transit service levels. If anything, the trend in Cluster 4 is pointing downward with San Bernardino, CA, which increased vehicle revenue miles by 28%, still losing 11% of ridership. Regions in cluster 5 exhibit no relationship between change in transit ridership and transit service levels. Boston cut service by 3% and increased ridership by 5.5%, while Washington D.C. increased service by 7.9% and lost 5% of ridership. As for small and medium-sized regions, large metro areas lost transit ridership systematically with an average drop of 1% in Cluster 4 and 4% in Cluster 5. However, unlike small and mid-sized regions, the change in transit service levels was not a significant factor of transit ridership change between 2012 and 2016 in large metro areas. Overall, Figure 11 shows that transit service levels were not responsible for the decline in ridership in mixed traffic modes between 2012 and 2016. In every cluster, an overwhelming majority of regions both increased transit service and lost ridership. The relationship between transit ridership and transit service levels for small transit agencies means that small transit agencies were able to minimize the
30 decline in ridership and in some cases even yield modest increases, but at the cost of increases in transit service. In larger regions, changes in transit ridership seem completely uncorrelated with increasing transit service levels.