Objectives for Setting Transfer Time Windows and Other Considerations for Transit Fare Policy

Transcription

1 Objectives for Setting Transfer Time Windows and Other Considerations for Transit Fare Policy 0 Resubmission Date: November 0 William Hui, P.Eng. South Coast British Columbia Transportation Authority (TransLink) 00-0 Kingsway Burnaby, BC CANADA VH N T: 0..0 F: 0.. E: William.Hui@TransLink.ca Word Count Abstract: words Text:, words Figures and Tables: (equivalent to,0 words) Total:, words

2 Hui i 0 0 ABSTRACT Transfers allow customers to board connecting services without the requirement to pay an additional fare to reach their destinations. As transfers are a core component of fare policy, it is important to select an appropriate transfer time window, defined as the time between the first and final boardings in a journey. The transfer time window selected is important to satisfy both revenue and customer needs. Different approaches can be utilized to derive transfer time windows. For example, a transfer time window can be derived by finding the trip possible with the longest duration on the transit network and setting a transfer time window that allows this trip to be taken on a single fare. Alternatively, a transfer time window can be set so that the time spent on the system is proportional to the fare paid. Unfortunately, both approaches yield relatively long transfer time windows if extreme travel scenarios are considered. To address this concern, an alternative approach was developed to consider more common trips. This analysis found that a transfer time window can be better derived by selecting commonly accessed destinations on the periphery of the transit network, calculating transfer time windows for each origin-destination pair for a weekday during the midday, and then taking the th percentile of the transfer time windows required for the origin-destination pairs identified in the sample. The transfer time window should then be rounded to the next - or 0-minute increment for communication purposes. Additional considerations when setting a transfer time window include revenue impacts, forward compatibility, congestion and delays, and enforcement. Finally, transfer time windows should be reviewed periodically as the transit system evolves to manage any revenue or customer risks due to changes in network design and travel patterns.

3 Hui INTRODUCTION Transfers are a common feature of transit systems as they serve as proof of payment as well as allow customers to access connecting services without paying another full fare. Transit agencies cannot provide direct services for all its customers and often require them to board multiple vehicles to reach their destinations. Transfers facilitate efficient transit network design for the operator while addressing customer needs by keeping transit fares low and legible. While transfers are necessary facilitate travel on transit, restrictions are instituted on transfer privileges to limit abuses of the transfer policy. These restrictions are necessary so that customers do not receive unlimited travel privileges for a single fare, which is a potential revenue risk for transit agencies. Common restrictions include a fixed time in which transfers are valid and limits on the direction of travel. The definition of transfer time window used in this paper is the time between first and last boarding. This is the time in which customers may board connecting services without paying another full fare. This is in contrast to the total travel time, which is the time between first boarding to final alighting. Typically, a customer can remain on the system as long as they make their final boarding within the allotted transfer time window. Therefore, the total travel time can exceed the transfer time window. Transit systems that institute transfer time window limitations often round them to the nearest half hour. Transfer time windows and the associated fare policies seem to be rooted in what the transit system has adopted in the past and not on existing network topology or fare structure. As transit systems expand in terms of geography or service hours, there is a need to review the transfer time that is provided as existing policies may no longer be adequate for the transit agency or its customers. Defining an appropriate transfer time is complex as it is necessary to balance the needs of transit customers with the financial needs of the operating agency. Using Metro Vancouver to illustrate examples related to the complexity of transfer times, this paper investigates the factors that should be considered when reviewing transfer times and provides some considerations on how this parameter can be defined by transit agencies. REVIEW OF OTHER AGENCIES Many transit agencies adopt a transfer policy for its operations, which typically involves a maximum transfer time. For this study, the transfer policies of North American transit systems were reviewed. A summary of this information is provided in TABLE. Note that commuter systems (such as GO Transit and Sound Transit) and any premium fares/express services have been omitted from the review as these services represent premium services and generally have different fare structures than conventional services. All fares represent 0 rates, with the exception of York Region Transit, which reflect prices in effect as of January, 0. Note that this paper focuses on transfer policy within a single, integrated, multi-modal transit system, and not transfer policies and/or agreements between multiple agencies.

4 Hui TABLE Transfer Time Windows of North American Transit Systems City Fare for Adult Single Ride Transfer Time Window (minutes) Atlanta (Metropolitan Atlanta Rapid Transit Authority) () Boston (Massachusetts Bay Transportation Authority) Directionality $.0 US transfers in -hour window Unidirectional Only available through Breeze Card Not available for cash purchases $. US (on CharlieCard) $.0 US (cash) () 0 Unidirectional Only transfer allowed () Brampton Transit $. () 0 Omnidirectional () Burlington Transit $.00 () 0 Omnidirectional () Calgary Transit () $. 0 Omnidirectional Chicago Transit Authority () $. US 0 Omnidirectional st transfer cents nd transfer free No transfers when paying cash Durham Region Transit $.0 0 Omnidirectional, but continuous trip (0) Edmonton Transit $. () 0 () Omnidirectional () System Halifax (Metro Transit) $. 0 Omnidirectional () Kelowna (BC Transit) $.00 0 Omnidirectional () Miami-Dade County Transit $.00 () 0 Unidirectional Cash users required to pay full fare on each boarding Omnidirectional () Mississauga Transit $.00 () 0 (Set from start of the route) () Montreal (Société de 0 () Unidirectional, continuous trip transport de Montréal) $.00 (Addfares apply at stations outside Montreal proper and the Airport) (0)

5 Hui City Fare for Adult Single Ride Transfer Time Window Directionality (minutes) New York $. US (on Metrocard) 0 Omnidirectional (Metropolitan Transportation Authority) () $.0 US (cash) Oakville Transit () $.00 0 Omnidirectional Ottawa (OC Transpo) $. () 0 Omnidirectional () Philadelphia (Southeastern Pennsylvania $.00 US Transfers available for $ for purchase on the service preceding the transfer. Transfers are for unidirectional travel only. Up to two transfers are permitted for each trip Transportation Transfers not permitted near the trip origin Authority) () Free transfer interchanges exist Portland (TriMet) () $.0 US ( zones) 0 (weekdays) Omnidirectional San Francisco Municipal Transportation Agency Seattle (King County Metro Transit) () $.0 US (All zones) $.00 US (excludes cable cars) () $. US (Off-peak all zones) $.0 US (Peak zone) $.00 US (Peak zone) 0 (weekends) 0 Omnidirectional (excludes cable cars) * Product called the -hour ticket Unspecified, but seems that King County offers time-based transfers that allow for omnidirectional travel privileges (0) Toronto Transit Commission $.00 () Generally unidirectional, continuous trip (no time component) Pilot program in place for time-based transfer on St. Clair () Vancouver (TransLink) $.0 ( zone) 0 Omnidirectional () $. ( zones) $.00 ( zones) Victoria (BC Transit) () $.0 0 Unidirectional Washington $.0 US (on SmarTrip) 0 Omnidirectional Metropolitan Area Transit Authority () $.0 US (cash) Winnipeg Transit () $.0 0 Omnidirectional York Region Transit () $.0 ( zone) $.0 ( zone) () 0 Omnidirectional

6 Hui TABLE shows that the transfer time window of North American transit agencies generally ranges between 0 and 0 minutes. Many systems adopt flat fares a single fare that applies to all customers for each customer type (adult, concession, etc.) regardless of how far they travel on the system. Most transit agencies reviewed offer omnidirectional (multidirectional) transfers. Such transfers allow unlimited transfer for customers to travel in any direction, including stopovers and round trips within the prescribed transfer time window. Compared to unidirectional transfer policies, omnidirectional transfers are simpler to enforce as fare inspectors do not need to assess if the customer is indeed traveling in a single direction. For some agencies that have adopted a smartcard technology, cash-paying customers (who typically pay more for the initial fare) are required to pay a new fare for each boarding. This policy promotes adoption of the smartcard technology, which reduces transaction costs for the transit agency. In general, it seems that transfers within the bus network are typically free of charge, but an additional charge is sometimes required for transfers to the rail network, particularly in American transit systems. OBJECTIVE OF TRANSFERS IN THE CONTEXT OF SINGLE FARES As transit agencies institute different transfer time windows and because transit agencies grow and evolve, it is worth investigating if there is an objective method to set appropriate transfer time windows. Changing this and other elements of transfer policies have far-reaching effects on how customers pay and travel on a given transit network yet little formal analysis is available on how transfer time windows are derived. This paper explores potential approaches. Transfers are only required in a system where all of the following are true: Single-use fares are available (passes generally allow for unlimited travel); Available services do not allow all travelers to travel directly from their point of origin to their intended destination, thereby requiring use of connecting services; and The service provider does not charge customers for each boarding. With a goal of facilitating mobility on the transit network for customers, the following objectives of transfer policies can be derived:. Allow the customer to travel from his/her origin to their intended destination on a single fare.. Allow the customer access to a specific quantity of service on a single fare. While both objectives can be achieved simultaneously, they may pose revenue and customer risks. Further expansion of the network may exacerbate these risks. As both objectives have benefits and drawbacks, it is worth exploring how they will affect the selection of a transfer time window. This paper will derive transfer time windows to meet the above objectives. Metro Vancouver will be used as a case study for this paper. While the analysis concentrates on extreme travel scenarios to illustrate outer limits of the potential transfer time window, in

7 Hui practice agencies will need to consider the extent to which they wish to accommodate extreme trips in their policies. Completing Trips on a Single Fare To achieve this objective, the time between the first and last boarding of the longest possible direct-routed trip in terms of duration can be made within the transfer time window. This is more practical for smaller systems so as to keep the transfer time window within reasonable limits. While this seems simple to achieve at the outset, there are several variables to consider when defining the longest trip in the system: Time of Day and Day of week Certain services operate on specific days of the week and operating frequencies sometimes differ depending on the time of day. Therefore, the time between the first and last boarding for the longest trip may differ between weekdays and weekends, and time of day. In some cases, the longest trip on weekdays may not be serviced by transit outside of peak hours or on weekends. Time of year While not as significant as the day of week or the time of day, the time of year is important to consider as certain services are seasonal, and frequencies of services may be adjusted throughout the year. Therefore, the time between the first and last boarding for the longest trip may differ throughout the year. Random transfers versus timed transfers This consideration is related to the time of day. The time between the first and last boarding will depend on connections through the trip, especially if one leg of the trip involves an infrequent route (other than the first leg). A trip relying on random transfers can take much longer than a trip with timed connections. For example, a customer relying on a random transfer may miss a connection for a bus that runs every hour. Therefore, to meet the objective of being able to complete this trip on a single fare, the transfer time window would need to be extended by one hour compared to the transfer time window that would be required if the customer made the timed connection. As such, the transfer time window may need to account for the frequency of connecting services and may substantially increase the transfer time window required. To illustrate the amount of time that is required to allow customers to travel from one end of the system to the other, an example will be drawn from Metro Vancouver using the following parameters, which are intended to represent the longest trip possible in the region: Day of week: Saturday will be used for analysis as service frequencies are lower on these days and are generally comparable to Sunday service. Sunday and holiday schedules are not used because some services do not run on Sundays, which in turn may limit the number of travel patterns available for analysis. Time of day: The analysis will be conducted for trips starting during the late morning or early afternoon. Although evenings may offer worst-case scenario bus frequencies in the system, some services do not run in the evening and therefore may limit the number of travel patterns that can be analysed. Furthermore, there is lower demand on the transit system during the

8 Hui 0 0 evenings, so it is less likely that a customer will make a long-distance transit trip during the evening than in the midday. Time of year: With the exception of seasonal service (such as services provided for school trips), services do not vary widely between seasons, therefore, analysis of service at any time of year can be done. Fall 0 has been chosen for this example. Transfers to connecting services: assuming travelers on long-distance trips will plan their trip ahead of time, timed transfers will be used. To determine the longest trip, locations on the perimeter of the service area were chosen as shown in FIGURE. Google Transit was used to compute the shortest travel time between each origin-destination pair in FIGURE using the parameters stated above. This tool was chosen because the algorithm automatically optimizes travel time by minimizing connecting time. This maintains the assumption that travelers will plan their trip and select the most efficient routing/itinerary. Furthermore, Google Transit was chosen because it shows itineraries that exceed three hours of travel time. Trips of this duration are not shown in TransLink s trip planner. Of all the selected origin-destination pairs shown in FIGURE, the trip requiring the longest transfer time window was from Ruskin to Boundary Bay. Using the parameters stated above, this trip requires a transfer time window of at least hours and minutes, the time between first and final boarding. The actual trip duration, the time between first boarding and final alighting, is hours and minutes. Therefore, if the objective of the transfer policy is to allow customers to complete any unidirectional, continuous trip on a single fare, the analysis indicates the transfer time window must be set to a minimum of minutes. This is more than three times the existing transfer time window given to customers in Metro Vancouver. If the transfer time window were to be increased to allow for this longest trip, many customers in the region would be able to make return trips on a single fare. In general, if this policy was adopted, customers can receive much more extensive travel privileges for a single fare than what is currently provided, which would have significant revenue impacts to the transit agency.

9 Hui 0 0 FIGURE Extremities of the Metro Vancouver transit network. Reviewing the results of the analysis, the following observations were made: The quality and number of connections is a more significant contributing factor to lengthy transfer time windows than distance. The longest trip by distance (Brunswick Beach to/from Aldergrove) does not require a lengthy transfer time window. Trips requiring lengthier transfer time windows are trips that begin and end with local, infrequent services, which are typically not conducive for timed transfers. The most direct route (minimizing distance travelled) may not be the most efficient in terms of travel time. For example, Google Transit routed the trip from Boundary Bay to Brunswick Beach via Surrey as the customer would have had to spend an inordinate time in layovers if they had chosen a more direct routing. This may not be known to customers, which may require transfer time windows to be increased if ridership is higher on services that are more direct. This consideration may also affect the assumption above pertaining to transfers to connecting services. Some trips in the analysis have benefited from relatively new services that connect town centers. Therefore, provision of direct or cross-town services may serve to reduce transfer time windows required by customers.

10 Hui Consideration should also be given to the ridership and frequency of such trips on the transit network. For example, the actual number of transit trips made between Ruskin and Boundary Bay should be taken into account. For a trip that takes hours on transit and 0 minutes via private automobile, it is anticipated that the number of customers opting to take transit for this specific journey would be limited. If no customers are making these trips, then there is no practical reason to provide such a lengthy transfer time window. Allowing Access to a Specific Quantity of Service This objective is derived from the view that customers who pay a single fare receive a travel benefit on the transit network for a specified period of time, regardless of origin, destination or routing. This is practical for larger systems, where a single fare may not provide sufficient time to travel from one end of the system to another. Therefore, giving all customers the same benefit provides a sense of equity with regards to access to the transit system. The primary issue with this objective is defining how much time should be given to customers for travel. The following sections will discuss how transfer time windows can be based on different quantities of service that are provided to the customer. In other words, this section explores how transfer time windows can be based on the cost of fares or the cost of service provision. Time Spent on the System When customers receive a transfer for a single fare, the transfer can entitle the customer to spend a specific, continuous amount of time on the transit system. Since transfer time windows are roughly the same for all customers, all customers can potentially derive the same benefit and access the same quantity of transit service. While this seems to address equity for customers, the ultimate goal for customers is not to simply have time on the system (analogous to a carnival ride), but to have mobility; to travel from one point to another. This raises policy questions regarding the use of a single fare: in a well-developed transit network, facilitation of long-distance travel will also enable return trips over shorter distances. All things equal, time may be considered as a proxy for distance the longer one remains on the system, the more distance is traveled regardless of travel direction. The amount of service consumed may be different in systems with different types of transit services. For example, buses travelling on urban streets, suburban streets and highways may cover different distances in the same amount of time. In larger cities, rail options may also be available, which cover even larger distances in the same timeframe. As such, it is possible to travel different distances by spending the same time on the system. As an example, in Metro Vancouver, average transit travel speeds vary from. km/h (. mph) on ferry to. km/h (. mph) on commuter rail. Furthermore, not all customers will have equal access to service as certain services are not available in all areas of the service region. Furthermore, the amount of time spent on the transit system depends on the network topology and service frequency. In certain areas, where services are primarily designed to serve the central business district, direct cross-town (suburb-to-suburb) services may be poor or unavailable. Therefore, these trips may involve circuitous routing, requiring the customer to spend more time on the transit system than if more direct services were available. Therefore, time is not always a good proxy for distance as the amount of service consumed can vary greatly depending on the service or combination of services used in a single trip. In multi-modal systems, setting a single transfer time window for all customers may not be

11 Hui equitable because different customers may be able to travel longer distances than others due to differing access to different modes. Cost Recovery This is perhaps one of the most basic methodologies to define the amount of benefits that should be provided. In many industries, the cost to customers for benefits is set such that it does not exceed the cost to the company providing the benefits. This ensures that the company does not lose revenue. To calculate a transfer time window based on cost recovery, operating statistics must be consulted. Using Metro Vancouver as an example, the operating cost per total vehicle kilometer in 00 was $. (). Assuming an average bus load of 0 revenue passengers, the operating cost per passenger kilometer is approximately $0.. Based on current fare levels (-zone), a single fare costs $.0. To recover operating costs, the fare would allow passengers to travel approximately. kilometers. Using the average speed of a bus (approximately km/h), the transfer time window should be set at approximately minutes. This is a very short transfer time window, but it should be noted that TransLink services are subsidized and that fare recovery ratio is not 00% as assumed in this methodology. Therefore, this example represents one outer limit of the potential transfer time window. Single Fares in Relationship to a Day Pass Many transit systems offer day passes, which allow customers unlimited access to the system over the course of a service day. The transfer time window can be set such that the ratio of travel privileges between a single-use fare and a day pass is roughly equal to the ratio of the costs between the two products. As an example, the cost of an adult day pass in Metro Vancouver is $.00 and allows unlimited travel, whereas the cost of a single-use adult fare with the same travel privileges is $.00. The ratio between the costs is approximately.. If the proportion of travel benefits were set to the same ratio, the -hour travel privilege for the day pass divided by the ratio of. would result in a transfer time window of hours and 0 minutes for single-use fares. It should be noted that fares purchased in bulk quantities are discounted. For example, the cost of 0 single-use fares offering the same travel privileges is $.00. This works out to $.0 per ride, a % discount from cash rates. In this case, a day pass is approximately. times the cost of a discounted fare. Performing the previous calculations using discounted fares results in a transfer time window of hours and minutes. It is recognized that most patrons using a day pass will not be using the transit system during the entire time in which the product is valid. It is more likely that most customers will travel between AM and midnight. Therefore, an -hour parameter should be used instead of a -hour parameter to account for this consideration. Using this parameter, the calculated transfer time window is hours and minutes, which is still more than five times the existing transfer time window. Transit agencies may choose to reduce the -hour parameter further to better reflect actual usage of day passes, which will result in shorter transfer time windows. While Metro Vancouver was used as an example, it should be noted that the above transfer time window calculation is dependent on the pricing of each fare product. Furthermore, the transfer time window is not always directly proportional to the amount of benefit received by customers. The amount of time spent on the transit system using a single-use fare differs depending on the origin and destination and is a consideration when adopting the above methodology.

12 Hui 0 0 A Moderate Approach The preceding analysis used scenarios that represented extreme travel behaviors, which resulted in equally extreme transfer time windows. For example, it is unlikely that many customers would travel from one end of the system to another on transit on a Saturday, or would use the day pass without leaving the system during an -hour period. As such, this section proposes a moderate approach to derive transfer time windows. This analysis examines the transfer time window required to travel between common destinations near the periphery of the network during the midday on weekdays. This is similar to the analysis conducted for the first objective, but this analysis addresses travel patterns that are more likely to occur: more trips start and end at common points of interest rather than the furthest extremes of the transit network. Weekdays are also chosen over Saturdays as it is more likely that customers will make long trips when services are more frequent and because longer distance transit trips are typically taken for work or school purposes. Using Metro Vancouver as an example, the more common points of interest near the periphery of the transit network are shown in FIGURE and the corresponding transfer time windows are shown in FIGURE.

13 Hui FIGURE Points of interest of the Metro Vancouver transit network.

14 Hui A B C D E F G H I J K L M Aldergrove Coquitlam Town Centre Haney Horseshoe Bay Langley Centre Lonsdale New Westminster Seymour Steveston Tsawwassen Ferry UBC White Rock YVR-Airport A Aldergrove B Coquitlam Town Centre C Haney D Horseshoe Bay E Langley Centre F Lonsdale - G New Westminster H Seymour I Steveston J Tsawwassen Ferry K UBC L White Rock M YVR-Airport - All required transfers can be made within existing 0-minute period All required transfers can be made within a 0-minute period All required transfers necessitate transfer time windows in excess of 0 minutes FIGURE Transfer time windows (minutes) between points of interest in the Metro Vancouver transit network.

15 Hui FIGURE shows that most trips within transit network can be accommodated within a 0-minute transfer time window. There are a small number of trips that require a transfer time window in excess of 0 minutes, and these are restricted to trips originating from or destined to Aldergrove or Haney. The trip requiring the longest transfer time window in this analysis is the trip from Haney to Tsawwassen Ferry, which requires a transfer time window of minutes or hours and minutes. Furthermore, the direction of travel also affects the required transfer time window as shown in FIGURE. Aldergrove (A) Coquitlam Town Centre (B) 0 Haney (C) Horseshoe Bay (D) 0 0 Langley Centre (E) Lonsdale (F) 0 0 New Westminster (G) Seymour (H) 0 Steveston (I) 0 Tsawwassen Ferry (J) UBC (K) 0 0 White Rock (L) YVR-Airport (M) 0 Outbound and inbound transfer time windows differ by minutes or less Outbound and inbound transfer time windows differ between and minutes Outbound and inbound transfer time windows differ in excess of minutes FIGURE Asymmetry of required transfer time windows (minutes). For example, the top-most cell indicates that the transfer time window required to travel between Aldergrove and Coquitlam Town Centre differs by minutes depending on the direction of travel. Most transfer time windows between points of interest in FIGURE do not depend on the direction of travel. The cells in the lightest shading typically have similar service frequencies at both ends of the trip. Journeys where transfer time windows differ between and minutes based on direction have more variation as they typically involve long journey times at one end of the trip. Transfer time windows that are affected most by direction of travel are trips that involve long journey times and infrequent services at one end of the trip.

16 Hui Of the origin-destination pairs shown in FIGURE, % of the trips can be completed within a 0-minute transfer time window. If this is adopted as fare policy, this value should be rounded to 0 minutes for communication purposes. The extra minutes can also serve as buffer to account for potential delays in the system, such as traffic congestion. Overall, this approach derives values for transfer time windows that are more consistent with current practice. The transfer time window allows customers using single-use fares to complete most long-distance trips and addresses revenue concerns by limiting the potential for round trips to be made using single-use fares. This approach also provides a basis for transfer time windows to be objectively derived and can be applied to North American transit systems. FURTHER CONSIDERATIONS Instituting transfer time windows in transit system can be complex. Impacts on the transit system must be considered holistically. Revenue Impacts If transfer privileges are abused, this will result in revenue losses for the transit agency. The following discusses how transfer privileges can be abused and how transfer time windows may be defined to limit the misuse of transfers. Customers Giving Transfers to Other Customers With paper transfers and other single-use tickets, customers are able to give or resell their transfers to other customers who would have otherwise been required to pay a new fare. This can be done repeatedly until the transfer time window expires. Each time this is done, the transit agency loses revenue equivalent to a single fare. As such, a smaller transfer time window will deter this type of abuse. Free Return Trips or Trip Chains A lengthy transfer time window may pose revenue risks as customers may be able to make additional trips (return trips or general trip chaining) without having to pay an extra fare.. As such, the transfer time window should be selected so as to accommodate customers making longdistance trips on the system by not forcing them to pay multiple times, while limiting the ability for customers making short trips to abuse transfer privileges. Designing for the Future When selecting a transfer time window, it is also important to consider its suitability for future operations. As cities grow, the transit network will change to accommodate that growth. As the transit system expands its service provision and the geographical coverage, the concept of the transfer time window should be reviewed as the transfer time window may become inadequate. A transfer time window may be too lengthy if comprehensive network changes are implemented in the transit network. For example, the introduction of regional connector services or the construction of rapid transit lines may significantly reduce travel times, thereby reducing the need for lengthy transfer time windows. The provision of more direct services would also lead to similar outcomes. Conversely, if the transit agency expands its service area using conventional transit, longer trips may be taken on the network, requiring longer transfer time windows.

17 Hui Buffer Time The analysis provided in this document reflects scheduled times. There are often extraneous factors that require buffer time to be added to transfer time windows. Two considerations are discussed below. Congestion and Delays Delays beyond what is accounted in service scheduling impacts transfer time windows as congestion and other factors may lead to missed connections. Depending on the frequency of the connecting services, a missed connection can result in significant wait times, which may require customers to purchase an additional fare to complete their journey. A buffer should be included in the transfer time window to account for these situations. Alternatively, operational and administrative policies may be implemented to address this issue. Specific Times for Departure/Arrival All transfer time windows researched through Google Transit represent the most efficient trips that can be made on transit. In most cases, this would require passengers to leave and/or arrive at their destinations at a specific time, which may or may not correspond to the actual time(s) that customers want to travel. For example, the most efficient trip may involve a transfer time window of 00 minutes, resulting in an arrival time of :0 PM. However, if the customer must arrive at his/her destination by : PM, he/she may need to take a trip that requires a transfer time window of 0 minutes because the connections are not optimized. Therefore, as customers may need to travel at sub-optimal, consideration should be given to including buffer time in the transfer time window. Issuance of Transfers Certain transit providers use paper transfers, which are issued by bus operators, electronic fareboxes, or vending machines. In some of these cases, transfer time windows that are shown are rounded to a certain time interval. As such, it should be recognized that in such cases, not all passengers will receive the same benefits in terms of transfer time window, although the difference in transfer time window received by different customers may be small. Operator Discretion Bus operators and station attendants understand that delays could occur on the system. If a bus operator is running behind schedule, they can exercise discretion if a customer s transfer expired shortly before they boarded the vehicle. In such cases, the driver may allow the customer on board, which results in the informal provision of a longer transfer time window than what was initially issued. Similar scenarios may occur at rail stations if customers can successfully plead their case to station attendants. While this is beneficial for the customer, this practice causes inconsistencies in the enforcement of fare policy, especially since operators exercise different levels of discretion. This may also introduce revenue risks if new transfers are issued in the process. Ridership and Frequency of Trips The analysis should also consider how often the longest trips are made on the transit network and how many customers make these trips. If these trips are not made frequently or if are only

18 Hui made by a small number of customers, then it may not be worthwhile to provide lengthy transfer time windows to accommodate these trips in consideration of the revenue risks involved. Enforcement Enforcement is a critical element in ensuring that the transfer policies are not abused. It is important to ensure that customers using the transit network have paid the proper fare and are not abusing transfer privileges. The absence of enforcement poses significant revenue and reputational risks to the transit agency. CONCLUSIONS Based on the preceding discussion, it was found that a transfer time window can be best derived by selecting common points of interest on the periphery of the transit network, calculating transfer time windows for each origin-destination pair by direction, and then taking the th percentile of the transfer time windows for those origin-destination pairs identified in the sample. This analysis should be done for weekday travel during the midday. This is a moderate approach which excludes trips that are not likely to occur on a frequent basis. While the analysis yields reasonable values for transfer time window, several considerations may require adjustments to the derived value. For example, to effectively communicate transfer time windows to the general public, the value may need to be rounded to the nearest or 0 minute increment. Furthermore, revenue impacts, forward compatibility, congestion and delays, and frequency of travel for the origin-destination pair are also considerations that should be taken into account when setting a transfer time window. Finally, transfer time windows should be reviewed periodically as the transit system evolves to address any revenue or customer risks. FURTHER WORK Further work may be done by applying the methodologies outlined in this paper to other transit systems. This would be useful to verify the applicability of the methodologies, especially in systems that have different network topographies, service frequencies and fare structures than those of Metro Vancouver (TransLink). Incorporating ridership and origin-destination data into the analysis may also prove useful as this may assist in the determination of specific travel patterns that should be considered in setting transfer time windows. A potential methodology would be to base transfer time windows on existing travel times on the network if the data were available. Furthermore, it should be recognized that the analysis presented in this paper relied on the selection of specific locations on the periphery of the network. If possible, a similar analysis using all stop points in the system will be more comprehensive and may yield other results or insights regarding transfer time windows.

19 Hui DISCLAIMER The opinions expressed in this article are those of the author, and they do not reflect in any way those of the South Coast British Columbia Transportation Authority (TransLink). ACKNOWLEDGEMENTS The author would like to thank Jeff Busby for his valuable comments and encouragement towards the publication of the paper, and for his practical advice on how to design transit to make the world a better place. Special thanks also go to Tina Robinson and Maria Su for their efforts in reviewing this paper.

20 Hui REFERENCES () Metropolitan Atlanta Rapid Transit Authority. MARTA Metropolitan Atlanta Rapid Transit Authority. Accessed November 0. () Massachusetts Bay Transportation Authority. MBTA.com > Fare and Pass Information for Bus Service. Accessed November 0. () Bowman, A. How much time do you have to transfer on the MBTA with a CharlieCard? CharlieCard. Accessed November 0. () Brampton Transit. City of Brampton Fares. Accessed November 0. () Brampton Transit. City of Brampton Transfers and Connecting Transit. Accessed November 0. () Burlington Transit. Fares. Accessed November 0. () Burlington Transit. -hour Transfer. Accessed November 0. () Calgary Transit. Calgary Transit: Cash Fares and Transfers. Accessed November 0. () Chicago Transit Authority. Fares CTA. Accessed November 0. (0) Durham Region Transit. Fares. Accessed November 0. () The City of Edmonton. Fares :: City of Edmonton. Accessed November 0. () The City of Edmonton. Types of Fares and Passes :: City of Edmonton. Accessed November 0. () The City of Edmonton. Transfer Policy :: City of Edmonton. Accessed November 0. () Halifax Regional Municipality. Tickets & Passes Metro Transit Halifax Regional Municipality. Accessed November 0.

21 Hui () BC Transit. Kelowna Transit Fares. Accessed November 0. () Miamidade.gov. Miami-Dade County Miami-Dade Transit Fares. Accessed November 0. () Mississauga. Mississauga.ca MiWay Bus Fares. Accessed November 0. () Mississauga. Mississauga.ca MiWay Transfers. Accessed November 0. () Transit Toronto. Mississauga Transit introduces time expired transfers Transit Toronto Newspaper Archive. Accessed November 0. (0) STM. Fare Information Montréal STM. Accessed June 0. () STM. STM Transfers. Accessed June 0. () MTA.info. mta.info Buses. Accessed November 0. () Oakville Transit. Oakville Transit: Fares + Policies. Accessed November 0. () OC Transpo. OC Transpo Tickets. Accessed November 0. () OC Transpo. OC Transpo Transfers. Accessed November 0. () Southeastern Pennsylvania Transportation Authority. SEPTA Cash, Tokens, Transfers & Zone Charges. Accessed November 0. () TriMet. TriMet: Fares (Tickets and Passes). Accessed November 0. () SFMTA Municipal Transportation Agency. Fare Details. Accessed November 0. () King County Metro Transit. Fares King County Metro Transit. Accessed November 0. (0) King County Metro Transit. Metro Tickets and Passes. Accessed November 0.

22 Hui () Toronto Transit Commission. TTC Prices. Accessed November 0. () Toronto Transit Commission. TTC Transfers. Accessed November 0. () TransLink. Single Fares. Accessed 0 June 0. () BC Transit. Transit Fares. Accessed November 0. () Washington Metropolitan Area Transit Authority. Metro Fares Metrobus. Accessed November 0. () Winnipeg. Transit Fares. Accessed November 0. () York Region Transit / Viva. Transfers Terms and Conditions. Accessed November 0. () York Region Transit / Viva. Prices YRT/Viva. Accessed November 0. () TransLink. TransLink 00 Annual Report. annual_reports/00.ashx. Accessed December 0.