Developing Guidelines for Evaluating, Selecting, and Implementing Suburban Transit Services (2006)

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.

20 CHAPTER 2: FINDINGS The purpose of this chapter is to organize that information and data into the findings of the research. To assist in the organization, three perspectives of the analysis of the information and data will be used: • Land-Use Assessment • Performance Measurement • Pearson Correlation Matrix These analyses will indicate whether the information and data can be aggregated into findings that appropriately generalize the results, or whether disaggregated use of the data and information may be of more value regarding other areas potential to use this research. LAND-USE ASSESSMENT In assessing the land-use conditions within the transit service areas, we considered “four D’s:” Density, Diversity, Design and Deterrents to driving. These measures were chosen in order to evaluate the level of transit-supportiveness of each service area. The methodology for calculating these indicators is described below. In order to calculate the measures, we first determined the service area of a given suburban transit service. In most cases, it was possible to obtain Geographic Information System (GIS) files of the routes directly from the transit provider. In other cases, we developed GIS files of the service areas based on published bus schedules. A few assumptions were built into the definition of service areas. In the case of fixed routes, we buffered the centerline of the route by ½ mile, which is the distance we considered to be walkable. For deviated fixed routes, we buffered the fixed route by the distance that the driver was allowed to deviate to pick up or drop off passengers. For example, in Minneapolis, one fixed-route service deviates up to ¾ mile based on passenger requests, and therefore we buffered ¾ of a mile around the fixed route. In the case of point deviation service, we included the entire zone that included the collection points and the area covered by the service. In the case of demand-responsive service, the entire service area was selected for study. Park-and-ride lots were assumed to have a 5 mile radius catchment area. Density The Density indicator was measured by calculating the number of people, households and employment in the study area. Data was most often available at the traffic analysis zones (TAZ) level provided by the metropolitan planning organization (MPO) in that region. In some cases, particularly for population and households, data was provided in different units of geography, such as census tracts. In order to calculate population, household and employment for each service area, we assumed that development was uniformly distributed throughout each zone. In GIS, zones were spliced based on the boundaries of the transit service area. The population, households and employment related to the zones and the fractions of zones that fell within the service area were summed to arrive at the service area totals. To determine population, household and employment densities, we then divided the totals for population, households and employment by the number of square miles contained by the

21 service area. This uniform measure enabled us to compare service areas within the regions and across the country. With the number of jobs and population per square mile, we were also able to calculate the service area’s jobs/housing ratio. Diversity To assess the Diversity of activities occurring in each service area, we evaluated the mix of industries and land uses present. Industry data on employment in each service area, where available from MPOs or other sources, was summarized and presented. Land-use data in GIS format was also obtained from MPOs, at times at the parcel level. We then determined which parcels were in the service area and summed the areas of all the parcels by land-use type. Some land-use categories were aggregated for sake of simplicity (e.g. wetlands, wooded areas, and parks all comprise open space). Once we determined the area for each land-use type, use we divided the subtotals by the total service area to determine the proportion of each land use. For detailed case studies, we reported these percentages. In cases where we did not have sufficient data for this level of detail, we reported the dominant land use. Design Design was measured in terms of sidewalk and street connectivity and whether the area would qualify as an “urban place.” Sidewalk connectivity was chosen as an indicator of the ability for pedestrians to walk to transit stops. This was evaluated on a scale of 1 to 5, with five indicating the highest level of sidewalk coverage. As shown in the chart below, the numerical measures are correlated with descriptions from the perspective of a pedestrian or a planner, depending on the training level of the rater. Table 2-1: Rating System for Sidewalk Coverage Rating System for Sidewalk Coverage For Planners For Laypersons 1 Most streets do not have sidewalks 1 A person cannot walk there; he/she must use the street 2 Many streets do not have sidewalks – there are many gaps in sidewalks 2 It is difficult to walk there – there are lots of gaps in the sidewalk 3 There are sidewalks on at least one side of most streets 3 A person could walk there but it would not be very easy or pleasant 4 There are sidewalks on nearly every street, but not always on both sides 4 It is fairly easy to walk there but there are some places where it could be improved (e.g. crosswalks, lighting needed) 5 There are sidewalks on both sides of nearly every street 5 It is very easy to walk there (extensive sidewalks, crosswalks, pedestrian crossing lights)

22 The street network was evaluated for its level of connectivity to determine whether transit riders would have options for direct routes to transit stops. This rating was also done on a scale of 1 to 5, with greatest connectivity being a 5. Images of sample street networks for each of the five levels were chosen to give raters a visual reference. Additionally, a text description characterizing each level of connectiveness was used as a guide, as shown below. The final element of the Design measures was evaluating whether a suburban service area included any places that could be characterized as urban in terms of development patterns, street space, or walkability. This was a yes/no evaluation of whether the study area has a place with buildings fronting on the street and defining a strong public space, such as a traditional “Main Street.” If the person evaluating the area could answer “yes” to all of the following questions, the area was determined to have a place with “urban” characteristics. • Does the service area include a place where most buildings are adjacent to the sidewalk, not set back from the sidewalk? • Does the service area include a place where there are few if any parking lots in front of buildings? • Does the service area include a place where there is high street wall continuity – a place where buildings are lined up next to each other with few gaps, providing a vibrant place for pedestrians to walk? Deterrents to Driving Deterrents to driving are characteristics of a service area that have the potential to encourage more people to choose transit over driving. We evaluated two measures: parking costs and transit priority features. Parking costs were defined in terms of average daily cost of off- street parking. If the study area included a place where free parking is generally not available, the value of this binary value was defined as “yes”. Transit priority features include traffic signal priority, queue jump lanes, exclusive transit lanes, or busways. The transit priority features measure was reported as either “yes” or “no” depending on whether the suburban transit service makes use of any of these features.

23 Table 2-2: Rating System for Street Connectivity Rating System for Street Connectivity Rating Description Aerial View 1 Very low level of street coverage; mostly a few collectors or arterials with a few cul-de-sacs 2 Cul-de-sacs and curvilinear roads predominate; there are few areas with grid coverage 3 Significant grid coverage but also a number of areas with cul-de-sacs/ dead ends 4 Extensive grid network with a few cul-de-sacs and dead ends 5 Complete grid network with no cul- de-sacs or dead ends

24 RELATIONSHIPS OF LAND-USE SERVICE AREA CHARACTERISTICS TO TRANSIT SERVICE AND PERFORMANCE In the section below, a series of graphics, first used at a presentation for the APTA Bus and Paratransit Conference in May 2005 at Columbus, Ohio, indicate the relationships and findings from a land-use assessment perspective. Figure 2-1: Research Objective

25 Figure 2-2: Typology of Services The routes that were analyzed for this portion of the report are indicated below: Figure 2-3: Case Study List of Services

26 The routes displayed the following characteristics: Figure 2-4: Spatial Adaptation Figure 2-5: Temporal Adaptation

27 Figure 2-6: Demand Level Some of the findings included: Figure 2-7: Moderate Density

28 Trip ends per square mile appears to be a different metric to use in this type of analysis. Figure 2-8: Fixedness and Productivity However, as will be discussed later, the local policy decisions appear often to accept this lower productivity as a trade-off for increased coverage. Figure 2-9: Productivity and Density

29 In this instance the best performing route was again the fixed route in Wilsonville. Also note that density in this analysis is “trip density” in contrast to the more usual population density analysis. Figure 2-10: Productivity and Land-Use Mix Logically having origins and destinations in one service area should result in improved productivity, although there are exceptions to this rule based on feeders to the regional network or concentrations of local residential area trips, such as by seniors or students. Figure 2-11: Service Level and Productivity

30 Overview of Results As indicated previously, in depth land-use data were able to be collected for four of the case study areas. In those instances the level of detail was significant, but the varied types of service delivery made aggregated comparisons extremely general, as indicated by the results described above. The reason that only four of the case studies were included the detailed land-use data was the lack of readily available data in a consistent format that could be similarly applied in the case studies. In general, it appeared that land-use data are becoming more readily available in many areas, but the lead agency for maintaining the data and the types of data maintained can vary from one locale to another. Further, although some transit operators are very familiar with these data, there are others that do not use the land-use information, especially in the specific ways developed in the research plan. As a result, there is no one path that can be prescribed as the technique to be employed in general to access similar land-use data across the country. In addition, our experience in trying to depict the detailed data as shown in the Service Area Characteristics Diversity pie charts was that a significant amount of time and therefore research budget were consumed in this effort. However, we would suggest that the general methodology employing the “Four Ds” as described in the research can provide comparative information at the local level that will assist in understanding the comparative potential of various land-use factors to better support suburban transit options. Further, that the peaks, ridges, points and plains terminology and analysis does accurately describe the best service delivery alternatives for a given disaggregated land-use area. It should also be noted that the majority of the effort being expended by transit agencies, as reflected by the types of services included in the case studies, involve trying to serve lower density areas with multiple land uses (residential, schools, some commercial and health care). The range of solutions, varying from fixed route to route deviation, do have some interesting land-use correlations as described above: • Most are operating in areas of less than 20,000 trip ends per square mile. This would appear to be a relatively new metric and perhaps a new threshold for transit agencies to consider in planning activities. • Trip density in a given area was not a consistent factor in attracting more riders per hour. • Land use with mixed development appeared to perform better than land use of one type, i.e. residential or commercial. Clearly in many instances land use dictated the types of services provided, such as the job access routes in suburban Detroit serving the industrial areas, thus circulators with direct connections to the work sites were the best fit. However, it was interesting to note that in Minneapolis the difference between the route and point deviation was the higher number of attractors that needed to be served, thus the stricter scheduling of route deviation, versus the flexibility of broader service area coverage with fewer attractors in the point deviation example. However, the productivity of the former was significantly higher than the latter. As will be discussed further below, these examples demonstrated several key findings that still appear to significantly influence suburban transit programs. First, there is a wide range of perspectives regarding the role of suburban services, with evidence that coverage is more important than productivity. Second, that recognizing the benefits of some of the coverage oriented programs has resulted in better working relationships with transit agencies and

31 communities, including passage of funding resources legislation. Conversely, in other areas locales have opted out of the transit district to make their own policy decisions and even provide funding for those services. Obviously, the ability to fund these services that have much lower productivity than many fixed-route systems is also critical to maintaining sustainability. PERFORMANCE MEASUREMENT Next more typical transit performance measurement analysis was performed with demographic, service delivery and pedestrian network factors evaluated for 20 case study routes using geographic information systems (GIS) software. Demographic data were obtained for the year 2000 from the Census Transportation Planning Package (CTPP). Street network files were obtained from the local metropolitan planning organizations (MPOs), except for Minneapolis, where Census TIGER files were used. Files showing fixed transit routes were obtained from the MPOs, when available, and were created in GIS when not available, based on route maps from the local transit agencies. The boundaries of deviated-route service areas were created in GIS, based on route maps from the local transit agencies. Route productivity data (passengers per revenue hour) were compiled from transit agency information gathered during the project case studies. The routes were characterized in two ways: (1) by the trip type served (the home end of a trip versus the work end of a trip) and (2) by the type of service (local fixed-route, flexible route, and commuter). The following table shows the routes that were evaluated and their characteristics: Table 2-3: Description of Case Study Routes Route Agency Type Trip End Margate A BCT Fixed Route Home Margate B BCT Fixed Route Home Margate C BCT Fixed Route Home Margate D BCT Fixed Route Home Cedar Mill Shuttle TriMet Dial-a-Ride Home 155 Sunnyside TriMet Fixed Route Home 156 Mather Rd TriMet Fixed Route Home 157 Happy Valley TriMet Fixed Route Home 204 Wilsonville Rd SMART (Wilsonville) Fixed Route Home 903 Federal Way King County Metro Deviated Route Home 914 Kent King County Metro Deviated Route Home 927 Issaquah-Sammamish King County Metro Deviated Route Home 421 Burnsville-Savage MVTA Deviated Route Home 152 Milwaukie TriMet Fixed Route Work 41 Hawthorn Farm TriMet Fixed Route Work 50 Cornell Oaks TriMet Fixed Route Work 201 Barbur SMART (Wilsonville) Commuter Work 1X Salem SMART (Wilsonville) Commuter Work 291 Redmond King County Metro Deviated Route Work 224 Shoreview-Roseville MVTA Fixed Route Work

32 Service Area Definition Each route’s service area was defined as follows: • Fixed Route—all areas within ¼ mile air distance of any branch of the route. • Dial-a-Ride—the dial-a-ride service area. • Deviated Route—the combination of the route deviation area and all other areas within ¼ mile of the fixed-route portion of the route. • Commuter—the areas within ¼ mile of the local service portion of the route, where customers would mainly be boarding in the morning. The destination ends of the routes (in both cases, transit centers) were not included. Demographic Factors The smallest geographic unit available—either Census block group or Census traffic analysis zone (TAZ)—was used in the evaluation. A weighted averaging process, based on the percentage of a census block or TAZ falling within a route’s service area, was used to calculate the following demographic variables, typically used in estimating the viability of transit service in a given area: • Population density—the number of persons per square mile within the service area. • Job density—the number of employees per square mile within the service area. • Percent of population 0-17 years old • Percent of population 65 or more years old • Percent of households with no vehicles available • Percent of employees with no vehicles available at home • Average median income—the median income was known for each census block or TAZ; a weighted average of these median incomes was determined for the service area as a whole. Service Delivery Factors The following service delivery variables were evaluated: • Adult peak fare—the case study agencies used a variety of fare systems, including the use of peak- and off-peak fares and zone-based fares. The lowest (e.g., one-zone) adult fare during peak periods was used. • Service area—calculated in square miles, using GIS software. • Weekday TLOS Indicator—The Florida Transit Level of Service (TLOS) IndicatorF1 measures a combination of service frequency and span. In this application, it measures the percentage of a weekday that locations within the service area have access to transit. It was calculated for fixed-route and deviated-route services as (# of weekday round trips) * (10-minute window of opportunity to access transit per trip) / (1,440 minutes per day). For the lone dial-a-ride service, it was calculated as (service span in minutes) / (1,440 minutes per day). 1 Ryus, Paul, Jon Ausman, Daniel Teaf, Marc Cooper, and Mark Knoblauch, “Development of Florida’s Transit Level of Service Indicator,” Transportation Research Record 1731, Transportation Research Board, National Research Council, Washington, DC (2000).

33 Pedestrian Network Factors The following factors relating to street network connectivity were evaluated: • Network Connectivity Factor—the number of links (i.e., street segments between intersections) within the service area, divided by the number of nodes (i.e., intersections). The index value ranges from about 1.7 for a well-connected grid pattern to approximately 1.2 for a cul-de-sac pattern.F2 • Average Minimum Circularity Ratio—the circularity ratio compares the area of a polygon to that of a circle with an identical perimeter and is calculated as (4 * pi * polygon area) / (perimeter ^ 2). The ratio ranges from 1 for a circle to 0.785 for a square to near 0 for long, thin polygons typical of blocks formed by freeways, railroads, canals, and other similar barriers to pedestrian travel. The circularity ratio was calculated for each block, the minimum circularity ratio was determined for all blocks falling within a given ½-mile grid square, and the average of the minimum circularity ratios was calculated based on all grid squares intersecting a route’s service area. • Average Block Size Factor—Florida defines 50 polygons (blocks) per square mile as the minimum needed to establish a multimodal transportation district.F3 The block size factor was calculated as the ratio of a block’s area (in square miles) to one- fiftieth of a square mile. An average value of 1.0 or less suggests a relatively dense, walkable street network. The average block size factor was calculated based on all blocks intersecting a route’s service area. PERFORMANCE FACTOR RELATIONSHIPS TO PRODUCTIVITY This section highlights the most promising relationships between the evaluated factors and route productivity. The six flexible-route services showed a strong correlation between population density and productivity. In the above example, there was more limited correlation to trip density. Among the local fixed-route services, the Broward County, Margate routes are in a cluster by themselves and will be further discussed below. The remaining local fixed-route services showed a fairly weak correlation between population density and productivity. 2 Ewing, Reid, Best Development Practices, APA Planners Press, Chicago, IL (1996). 3 Florida Department of Transportation, Systems Planning Office, Multimodal Transportation Districts and Areawide Quality of Service Handbook, Tallahassee, FL (November 2003).

34 Figure 2-12: Population Density versus Productivity R2 = 0.9514 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Productivity (passengers/revenue hour) P op u la ti on D en si ty ( pe rs on s/ sq u ar e m ile ) Fixed Route Flexible Route Commuter Linear (Flexible Route) MVTA 224 Margate Portland/Wilsonville TriMet 41 There was some correlation between the productivity of the employer-oriented services and the percentage of employees that had no vehicle at home: Figure 2-13: Zero-Car Households versus Productivity R2 = 0.3854 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Productivity (passengers/revenue hour) P er ce n t Jo bs w /N o C ar Home Work Linear (Work) TriMet Cedar Mill Shuttle

35 There was some correlation between productivity and the amount of service provided, as measured by the Florida TLOS Indicator, which includes both the span and frequency of service: Figure 2-14: TLOS Indictor versus Productivity R2 = 0.3318 R2 = 0.3855 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Productivity (passengers/revenue hour) TL O S In di ca to r V al u e (% o f D ay w it h S er vi ce ) Cedar Mill Shuttle excluded There was relatively good correlation between productivity and the service area size with the logical result that the larger the service area, the less productive the service: Figure 2-15: Service Area versus Productivity R2 = 0.4086 R2 = 0.4945 0.000 1.000 2.000 3.000 4.000 5.000 6.000 7.000 8.000 9.000 10.000 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Productivity (passengers/revenue hour) Se rv ic e A re a (s qu ar e m ile s)

36 Factors that showed no apparent correlation included fare, percent of population under 18, and all three of the walkability factors. It should be noted that all of these comparisons are based on single-variable regressions; multiple linear regression and other more advanced statistical techniques might reveal other trends in the data. However, given the variety of services studied and the differences between geographic areas and land uses, the potential for productivity trends for aggregated service data appear to be limited. The demand-responsive services, when looked at as a group, tended to show better correlation for several factors that showed some correlation for all services combined. Two possible explanations for this are: (1) the demand-responsive services tended to serve larger areas than the fixed-route and commuter services, and (2) none of the demand-responsive services overlapped with each other. Overview of Results This performance measurement analysis also disclosed some other factors that would influence the results. Such as, when a ¼-mile service area around a bus route is assumed as the service area, it normally only covers a portion of individual block groups or TAZs in suburban areas, as block groups and TAZs tend to cover relatively large areas outside of central business districts. As a result, there can be significant variations in population density and other factors within a block group: multi-family residential might be located along the arterial street served by transit, with single-family residential located farther away from the street. In addition, larger TAZs in more suburban area would tend also to potentially skew the display of land-use data. For example, if one portion of a TAZ contained a relatively dense development, yet there were no other developments in the TAZ the overall area would appear to have a low population density. These variations cannot be captured from the information available from the Census and would require more detailed land-use information, which is often available from MPOs. However, experience would suggest that transit agencies typically use TAZ data, thus another lack of consistent connection between land use and transit data. Demand-responsive services, in contrast, have a higher likelihood of serving entire census blocks and TAZs, and thus variations in density within individual census blocks/TAZs are blended out and comparisons between services in different cities tend to be more equal. The results from Broward County, which aggregated results to the city level, showed better correlations than the route-by-route results presented here and therefore suggest that evaluating larger areas may be more appropriate when trying to decide what type(s) of suburban services are appropriate. Many of the fixed-route services that were studied (Margate, Portland, and Wilsonville) had service areas that significantly overlapped with other fixed-route services. Because the overlapping services covered areas with relatively similar population densities, any differences in productivity would be the result of other factors. In contrast, all of the demand-responsive services that were studied served unique areas that were not part of the service areas of other studied routes. Thus, the variety of services that were included in this analysis from various parts of the country did not provide many significant findings with the possible general information: • Use of population, not trip, density did prove to have a good correlation, especially for demand-responsive services.

37 • The size of TAZs in suburban areas may impact results from this type of performance analysis based on the potential impacts on the available data for the variables used in the analysis. CORRELATION MATRIX The final performance indicator used to assess the findings was the correlation matrix from the Broward County including Margate and surrounding areas as discussed below: The Correlation between Passengers per Revenue Hour and Transit Utilization Factors The hypothesized relationships between passengers per revenue hour and such measures as population density, income, the elderly segment of the population, the student-age segment of the population, the number of owner-occupied units, the number of renter-occupied units and car ownership were tested at the route level with data derived from the census blocks which permitted Pearson correlations to be conducted to measure the magnitude and sign of these relationships. Pearson's correlation reflects the degree of linear relationship between two variables. It ranges from +1 to -1. A correlation of +1 means that there is a perfect positive linear relationship between variables. It is a positive relationship because high scores on the X-axis are associated with high scores on the Y-axis. A correlation of -1 means that there is a perfect negative linear relationship between variables. It is a negative relationship because high scores on the X-axis are associated with low scores on the Y-axis. A correlation of 0 means there is no linear relationship between the two variablesF4. The correlation between passengers per revenue hour and income shows clearly that as the level of income declines the passengers per revenue hour rises and this noticeable inverse relationship confirms standard transit utilization theory. The elderly and student age segment are both positively correlated to passengers per revenue hour which also confirms transit utilization, though in this sample set, the relationship is rather minimal to non-significant. However, it is mildly interesting to note that the correlation between student population and transit ridership is stronger than the correlation between elderly and transit utilization. Population density is highly positively correlated to passengers per revenue hour in the routes analyzed, so standard transit utilization theory holds firmly in this local circulator setting as well. Owner-occupied housing units had a mild negative correlation to passengers per revenue hour, showing that as the number of owners rose along the routes examined, it is expected that ridership per hour would decline. The number of renter occupied units was slightly positively correlated with passengers per revenue hour though the magnitude of this relationship is too small to be considered a strong factor. The segment of owner occupied units with no car was strongly correlated to passengers per revenue hour. This finding is again consistent with standard transit utilization theory. The segment of owner-occupied units with one car is also positively correlated with passengers per revenue hour. This might be due to the fact that the owner occupied households with only one car have more people in the household with mobility needs that are not being met with a single car. The segment of renter occupied units with no car is positively correlated with passenger per revenue hour, once again consistent with the notion that the absence of personal transportation, especially in the case of persons renting units, implies transit utilization for many 4 Hyperstat Online Contents, “Pearson’s Correlation,” http://davidmlane.com/hyperstat/A62891.html

38 trip purposes. The segment of renter occupied units with one car is slightly negatively correlated with passengers per revenue hour, so as renters get personal vehicles, ridership on the shuttle system would decline. This finding might reflect that renter occupied units have fewer people and less travel demand. The chart and correlation matrix below show the results from the statistical analysis. Table 2-4: Broward County Correlation Matrix Pass. Rev. Hr./Income -0.57648 Pass. Rev. Hr./Elderly Segment 0.061163 Pass. Rev. Hr./Student Segment 0.090209 Pass. Rev. Hr./Population Density 0.83333 Pass. Rev. Hr./Owner-Occupied -0.39667 Pass. Rev. Hr./Renter-Occupied 0.036481 Pass. Rev. Hr./Owner-Occupied No Car 0.694742 Pass. Rev. Hr./Owner-Occupied 1 Car 0.380401 Pass. Rev. Hr./Renter-Occupied No Car 0.520486 Pass. Rev. Hr./Renter-Occupied 1 Car -0.12368 Demographics It might seem obvious to many that certain demographic characteristics contribute to better transit ridership, but with such limited experience in the provision of local circulators in primarily suburban settings, it was of benefit to confirm if normal indicators of transit potential apply to local circulators as they do to regular fixed-route transit service in a more regional setting. As noted above, there is a very strong (0.833) positive relationship between transit use and population density for the local circulators that were studied. In short, the higher the density, the higher the transit ridership per hour was for the local circulators. Not too far behind in terms of relationships was the high positive correlation between lack of car ownership and transit use. Perhaps a little surprising was that the relationship was even stronger for owner-occupied dwellings (0.69) without cars versus renter-occupied dwellings (0.52) without cars and transit ridership per hour. It is hypothesized that rental apartments are usually smaller than owner-occupied homes, and there might be more total need for mobility in an owner-occupied dwelling due to more people living in the owner-occupied home. As expected, there was also a strong negative correlation (-0.58) between income and transit ridership per hour. In other words, the higher the income, the lower transit ridership per hour was in the local circulator systems. Thus it was logical that transit ridership per hour was most successful in the City of Lauderhill. While the average passengers per hour for all five routes in Lauderhill were 22.0, two of the routes came close to carrying 30 passengers per hour. In Lauderhill, the population per square mile is 8,179, easily the highest among the eight city circulator systems reviewed. The median household income is $32,070, which is among the lowest of all eight city systems reviewed. Lack of car ownership (9.9% of the owner-occupied dwellings and 12.1% of the renter- occupied dwellings) was above average, but not extraordinarily so. It is interesting to note that the City of Dania service area has virtually the same median household income as Lauderhill’s service area, a similar percentage of renter households without cars, and a better headway (40

39 minutes versus 45) than most of the Lauderhill routes. However, Dania’s population density is only 3,272 persons per square mile, and realizes a local transit circulator ridership per mile that is slightly less than one-third that of Lauderhill. In fact, the listing of cities in order of passengers per mile follows quite closely to the listing of the cities by their population densities, regardless of other demographic characteristics. Overview of Results The first two evaluation techniques had a broader universe of case studies while the third focused on data only from Broward County, which resulted in a more uniform set of findings. As indicated above, within the parameters of local circulator funding which has been established by the local transit agency the correlations are similar to those typically found within transit analysis, such as: • Higher population density results in higher transit usage • The combination of higher population density and lower income also increases transit use • To a lesser degree, lack of auto ownership, especially by residents, increases potential transit use • In this instance, youth were more likely to use transit than the elderly Thus, with the exception of the last finding, the results are more consistent with typical transit analysis. This indicates that measuring similar services within a given geographic area would likely lead to more specific findings. SUMMARY AND CONCLUSIONS There were limited results of value from the use of various tools to analyze the data from the detailed case studies to determine significant trends that could be reported from an aggregated data perspective. However, there were consistent results recorded when data were collected in a specific local area. In addition, there appeared to be a number of data collection and display consistency issues resulting from availability and scale differences.