Principles of Paratransit System Design
Transportation Alternatives is the pioneer in identifying and documenting the factors which affect the efficiency or performance of paratransit operations. Many of the principles outlined below were initially developed during TA's USDOT-funded nationwide examination of special paratransit systems (1978-1980) which resulted in the Department's publication of the three-volume manual on paratransit performance, decision-making and system design — the project's primary work product — authored by Mr. Einstein. This project's findings included a number of revolutionary notions, to which many of Today's ADA-era complementary paratransit systems have still never adopted.
Factors which do not Affect Performance
One of the most startling observations to emerge from this analysis was the litany of factors which, by themselves, have either no effect at all, or relatively minor impacts, on performance. These factors include:
- Dispatching skill
- Scheduling skill
- Service area size
- System user density
- Density of origins
- Density of destinations
- Ambulatory status of passengers (i.e., wheelchair occupancy or walking aids)
- Labor rates
- Level changes accommodated (door-to-door versus curb-to-curb)
- Vehicle size and type
- Fares and farebox recovery ratio (i.e., operating ratio)
- Lead or Operating Agency type(s) and relationship
- Trip Lengths and Ride Times
- Reliability and Wait Times
How can this possibly be?! The answer lies in an examination of the factors which do matter — and why they matter. Before examining these factors, a comparison of two pre-ADA dial-a-ride systems illustrates how unimportant the factors listed above truly are.1
Presented in the table below are a few salient characteristics of two pre-ADA paratransit systems:
|Service Area Size (sq. miles)||22||50|
|Eligibility||disabled||elderly and disabled|
|Consolidation||single system||separate for Native Americans|
|Average Ride Time (minutes)||27||30|
Apart from the performance (in terms of passenger trips/hour) identified here, it would seem only logical that a consolidated system in a small, dense urban service area with a lot of vehicles (e.g., Boston) would certainly be more productive than a fragmented one in a bigger, largely-rural service area with only a handful of vehicles (e.g., Tulsa). Yet not only was this not the case, actual performance was in dramatic contrast to it. What could possibly explain this? (The greater density of elderly riders in Tulsa compared to only the disabled riders in Boston does not.) The answer belies the depth of mythology about paratransit - and helps to explain why so many of Today's ostensibly modern (and largely digitally-scheduled and dispatched) paratransit systems remain deplorably inefficient, despite a dazzling array of technologies available to articulate practically every nuance of operations and even overcome limited technical knowledge at the operating level.
Factors which Matter
Picking up on the Boston/Tulsa comparison, a brief glimpse of these systems' designs, ridership policies and operating parameters documents the remarkable distinction between them in a few key areas.
Boston. In Boston's "The Ride," passengers were permitted to ride wherever and whenever they wanted to within the service area to the degree space was available. (Often trips that could not be accommodated when requested were "negotiated" to different time slots.) Most service was provided on either a pre-scheduled demand-responsive or immediate-response basis. Because the bulk of service was allocated largely for medical purposes, the return portions of even most pre-scheduled trips were "dispatched" (i.e., provided on an immediate-response basis). Because this practice was allowed in the pre-ADA era, schedules were "filled" on a first-come/first-served basis.
Tulsa. Despite the array of factors which would superficially generate a far lower degree of performance - including separate transportation of Native Americans from the rest of the service area population2 — Tulsa's system was organized by a strict set of principles which stratified rides according to pre-scheduling format, trip type and vehicle size and type - and reinforced them with fares whose equity reflected this stratification:
|Trip Purpose||Pre-Scheduling Format||Vehicle Type/Size||Fare|
|Shopping||Subscription||Full Size Transit Bus||25 cents|
|Work||Subscription||Full Size Transit Bus||25 cents|
|School||Subscription||Full Size Transit Bus||25 cents|
|Social Service Agency/Nutrition||Subscription||Full Size Transit Bus||25 cents|
This stratification reflected a trip-sorting logic that respected which times of the day certain trip types were needed versus those trips which could be provided virtually any time during the day. Thus, typical peak hours were reserved largely for work and school trips, and for those medical trips which (as an exception to the scheme outlined above) reoccurred regularly (dialysis, chemotherapy, radiation, physical therapy) and could, therefore, be provided on a subscription basis. Pre-scheduled demand-responsive and immediate-response trips were inserted into these "skeletons" where they could be accommodated, making the schedules even more productive.
Off-peak service was reserved primarily for other trip types — although even these trips were provided in a framework that respected temporal patterns of need. For example, trips to nutrition program sites were provided between 11:00 A.M. and noon (outbound) and 1:30 and 2:30 P.M. (return trips). Again, other trips were inserted between this noontime framework and peak hour trips.
These temporal considerations were used to optimize performance in the schedule-building process. A productive schedule between 8:00 A.M. and 5:00 P.M. (or 08:00 and 17:00 in the "military time" parlance commonly used in paratransit operations), provided on full-size transit buses, might include the following primary trip types:
- 08:00 - 09:30: work and school, front end
- 09:30 - 10:30: medical/social-recreational trips, front end
- 10:30 - 11:00: grocery shopping/supermarkets (morning runs), front end
- 11:00 - 12:00: nutrition, front end
- 12:00 - 12:45: grocery shopping/supermarkets (morning runs), back end
- 12:45 - 13:30: grocery shopping/supermarkets (afternoon runs), front end
- 13:30 - 14:30: nutrition, back end
- 14:30 - 15:30: grocery shopping/supermarkets (afternoon runs), back end
- 15:30 - 17:00: work and school, back end
Within this framework, a bus driver might pick up and drop off a vehicle practically full of grocery shoppers from one or two major origins (e.g., senior citizens housing complexes) and, during the hour in which they shopped, would provide other group trips (e.g., nutrition trips). In contrast, a van driver would very likely provide either isolated demand-responsive trips or small groups of them would very likely provide either isolated demand-responsive trips or small groups of them (e.g., for medical or social/recreational purposes) during this interlude. Regardless, the "core" of service was scheduled around highly-efficient subscription trips, filling in the "gaps" with pre-scheduled demand-responsive and immediate-response trips.
Principles of Paratransit Performance
The pre-ADA Tulsa Dial-A-Ride (circa 1978) illustrates how the optimization of several of the principles of performance leads to extraordinary performance even when other variables are not optimized. Of the six factors which affect performance, the Tulsa system optimized only four (bolded below)3:
- Pre-scheduling Format
- Type of Trip
- Fleet-to-User Density
- Service Cost or Contract Rate
- Service Concept
With six vehicles, a large service area and fragmented service, Tulsa's fleet-to-user density was extremely thin. Operated by its transit agency, often with full-size transit buses, its service costs were relatively high. Yet the transit agency was so creative, systematic and enlightened in its application of the other four factors that mattered when optimally structured (remember: this was 1978!) that it achieved remarkable performance despite the overall service area's extremely thin fleet-to-user density. An examination of these six factors, and their relationship to one another as well as other system variables, demonstrates why they are so critical to performance.
Pre-Scheduling Format. In terms of trip scheduling, there are only three formats, all dependent on the degree to which each trip meets two criteria:
- Can the trip be pre-scheduled?
- Does the trip reoccur regularly?
Illustrating the answers to these questions diagrammatically, and assigning names to each classification among them, one can derive the three trip formats:
|Is the trip regularly reoccuring?|
|Can the trip be pre-scheduled?||Yes||Subscription||Pre-Scheduled Demand-Responsive|
(Obviously, there is no such thing as a regularly-reoccurring trip that cannot be pre-scheduled.)
Because there is more time to manipulate and "negotiate" them, and because certain trip types lend themselves more to certain of these pre-scheduling formats, subscription trips are more productive than pre-scheduled demand-responsive trips, which are more productive than immediate-response trips. More importantly, a schedule properly combining all three trip types is more productive than one composed of one or two of the three (taxicab trips are dispatched only as immediate-response trips, even though many of them are pre-scheduled at the trip reservation level) — and a logical schedule-building approach (a la Tulsa) incorporates this logic. For example, a core of subscription trips may serve as a "skeleton" into which demand-responsive trips are inserted. Moreover, a schedule comprised largely of subscription trips can be continually optimized such that, over time, it evolves in efficiency: Fewer trips per day have to be scheduled, and focusing on their origin and destination pairings provides "reservation clerks" with an opportunity to "negotiate" the consolidation of a pair of such trips into the same run by the same vehicle. The operating principle which emanates from this approach is that every schedule change is an opportunity for performance improvement.
Type of Trip. The mix of trip types (or trip purposes served) is critical to performance because:
- the density of each type of destination is different
- users’ trip needs cannot be satisfied to the same degree by the "nearest" destination of each trip type
As an example, doctors' offices, medical centers, clinics and out-patient hospital services may greatly outnumber supermarkets. However, except in rare cases, most individuals cannot use the "nearest" doctor, whereas their grocery shopping needs can generally be met at the nearest supermarket. So a schedule comprised heavily of shopping trips will generally be more efficient than one consisting largely of medical trips. This efficiency is further reinforced by the fact that it is far easier to group shopping trips since they are not as "time-sensitive" as medical trips (i.e., one rarely needs to make an appointment to shop whereas one cannot predict illness).And, of course, the density of paratransit riders who regularly take shopping trips via paratransit (e.g., ambulatory elderly individuals) is generally greater than that of those who regularly take medical trips (e.g., ambulatory elderly individuals) is generally greater than that of those who regularly take medical trips (e.g., disabled elderly and non-elderly individuals)
Eligibility. Eligibility is strongly related to performance because different population sub-groups take different types of trips, and concomitantly, certain trip types lend themselves more favorably to certain performance-enhancing pre-scheduling formats than do others. Beyond the relationship to trip types, non-disabled elderly persons can often take even the same types of trips as their disabled counterparts on a more "performance-friendly" pre-scheduled format — although there are many exceptions (a dialysis, chemotherapy or radiation trip can be provided on a subscription basis whereas an illness-related doctor's visit obviously cannot). Further, because of the difficulties involved, certain ridership populations tend to not use paratransit services for certain trip types — for example, wheelchair occupants often have friends or relatives take them grocery shopping, -or do it for them (whereas these "aides" cannot take their medical treatments for them).
Fleet-to-User Density. As the initial example (see the first table of bullet points above) so well illustrates, neither fleet nor user density, by itself, is critical to performance. However, the combination is extremely critical to it. All other things equal, a thick fleet and user density will generally yield far higher performance than thin densities of either or both. This is simply because more riders can be picked up and dropped off when their origins and destinations lie more closely together. Factors like traffic offset this relationship to a degree, but where traffic is greatest, the combined fleet-to-user density is generally thick enough to more than offset it. And this principle does not even consider the deployment of exclusive ride services, like taxicab service, to provide trips more difficult to group. (The increasing inclusion of low-floor, ramp-equipped minivans as a replacement for taxi sedans has also broadened this capability to provide trips to wheelchair users as well as ambulatory and semi-ambulatory passengers.)
Service Cost or Contract Rate. Obviously, how much service costs affects certain types of performance measures related to these costs, as measured by indicators such as "cost per passenger trip." One might also surmise (incorrectly) that paratransit providers with high costs would be more cost-conscious than those experiencing lower costs. Such is clearly not the case. This is partly because the delivery of service in costly urban areas is more complex, and partly because urban areas rarely exercise the creativity in "service concept" design that rural and suburban areas employ — where trip lengths are much longer, and where costs would escalate severely were no service concept employed. Of course, there are exceptions to this rule, particularly where digital scheduling technology has been employed in place of intelligible "system design," and as a result, the articulation of the technology merely exaggerates the chaos which the lack of intelligible system design would have provided as a foundation.
To blame the absence of a service concept on the Americans with Disabilities Act--merely because eligible paratransit riders may travel wherever they wish within the transit service area - simply reflects a misunderstanding about an important nuance of ADA paratransit requirements and a lack of imagination in response to it. This is so largely because the ADA permits municipalities and transit agencies to differentiate fares with respect to pre-scheduling formats — and even suggests this approach in the provisions of the ADA regulations themselves. And, of course, there are no prohibitions against a community minimizing trip lengths by using zonal structures - so long as the fares for any zone-to-zone trip do not violate the double-the-bus-fare "ceiling" of the ADA for fixed route trips of the same length.
Service Concept. Simply defined, a service concept represents the deployment of vehicles in time and space. More than any other factor, a clever service concept enables a system to overcome the constraints of other performance-related factors—particularly those related to density. While the service concepts possible are limited only by one's imagination, experience and knowledge, a single example illustrates how productive a well-conceived concept can be.
In pre-ADA Portland, Maine, the paratransit system provided 3.1 trips per hour within a 300 square mile service area with only 10 vehicles. How was this possible? The service area was divided into an inner core (City of Portland) and an outer ring surrounding it, which was divided into ten "sectors." Drivers were literally hired by sector so that they could "park out" their vehicles at home (in their respective sectors) — effectively minimizing A.M.-starting and P.M.-ending "deadhead" time and mileage. Early morning service began with subscription trips for work, school and selected medical purposes (dialysis, chemotherapy, radiation, etc.) provided as vehicles drove from the outer edges of their "sectors" toward the urban core (where these destinations lay). Once within the core, most vehicles remained there and provided service in much the same manner as did Tulsa (without the differentiation of vehicles and fares). P.M. work, school and selected medical trips were, similarly, provided as outbound trips to the vehicles' decentralized "storage locations" (i.e., "park-outs") in the outlying sectors.
Compared to fixed route transit, paratransit services are less a captive of the urban form in which they operate. And because they respond dynamically to travel patterns (rather than being designed to reflect them on a long-term basis), paratransit services can be shaped to both accommodate and manipulate user needs in accordance with principles of system design which optimize performance.
Mythology, Accountability and the Cost of Failure
Reflecting upon the principles outlined above, it is no surprise that the substitution of computerized for manual scheduling and dispatching has not improved performance in so many systems — and in many cases has undermined it. For starters, most paratransit trips should not be dispatched to begin with — with the exception of dense urban areas with large taxi fleets (and, thus, high fleet-to-user densities), where trip-grouping often does not pay off. More importantly, software can only optimize performance within the context of the system, as it is designed. Otherwise, while a software program can schedule or dispatch trips better than a marginal scheduler or dispatcher, its superiority over a competent scheduler will hardly be significant enough to compensate for a poorly-designed or non-existence service concept. The greatest benefit of computerization lies in the fact that certain operating functions can be performed with precision by management personnel with less than perfect knowledge of these functions. Also, much larger systems can be operated without overwhelming reservation, scheduling and (particularly) dispatching personnel. And service reliability and equity can be far more uniform from system to system, and can more easily withstand personnel turnover and other upheaval.
It is also important to recognize, of course, that computerization of operating functions has become practically essential to most large service areas which must now respond to the numerous requirements of the ADA. However, it is critical to recognize that the transition from analog to digital paratransit operations in and of itself has not, will not and cannot lead to significant performance improvement. Thankfully, it can enable communities to articulate the operating functions of the service concepts they define, and help optimize the performance-enhancing manipulation of the variables which do indeed matter.
It is also worth noting that "real time" or immediate-response service is not necessarily a more valuable or "higher" form of service — even if it does provide the freedom of choice guaranteed by the ADA. For those trips which can be accommodated by it, subscription service is actually a more valuable form of service: Riders do not have to call to reserve every trip but, instead, vehicles arrive automatically and consistently, and riders need call only to cancel individual trips. Offering subscription service also decreases "no-shows": Riders who abuse the privilege are simply suspended from the subscription roster. The charging of lower (or no) fares for subscription (or any pre-scheduled) service, which the ADA permits, (see page v-II in the introductory section of the ADA Paratransit Handbook accompanying the ADA), further reinforces rider cooperation and minimizes "no-shows."
These relationships underscore yet another critical principle of paratransit operations: Many characteristics which benefit a system (e.g., pre-scheduling formats which improve performance) also benefit its riders. Following this is the fact that the better a system seems to be for its users, the less it seems to cost - a relationship the examples above corroborate forcefully. (This caveat also factors in the costs of liability exposure and regulatory non-compliance which often accompany poorly- or non-designed systems, particularly when vehicles run behind schedule, and drivers cut corners in a variety of areas that compromise passenger safety.) In simple terms, it costs less to operate a productive and imaginative system than a poor one.
Ultimately, a system's performance reflects its design — not the manner in which that design is articulated. The failure to grasp and respond to this principle has proven costly to both paratransit systems and their riders, not the mention the tens of billions of dollars wasting, as a national matter, over the course of the ADA, where the principles outlined above almost universally ignored and violated.
1 Pre-ADA systems are particularly valuable for such comparisons because they are not skewed by factors unrelated to the time-and-space relationships which govern paratransit performance—such as the coincidence of a potential system user's origin and/or destination to a fixed route line, which effectively screens out swaths (which exist between fixed route bus corridors) of potential users and skews notions of density to which meaningful performance variables are related. Pre-ADA systems which often transported both elderly and disabled individuals are particularly useful for performance-related comparative purposes.
2 This anomaly was not the result of racism but rather, the fact that origin and destination pairings for these two groups were not coincident and, further, that each group relied on different taxicab and van providers within their respective sub-service areas as a matter of choice. Regardless, this stratification has long since disappeared. It is referenced here only as it helps to illustrate certain principles of paratransit system performance.
3 One of the factors which matters—service costs or contract rates—was not relevant to this example, which measured performance in terms of passenger trips per hour. However, the transit system providing this service was the unionized, public sector transit agency, often using large buses costly to purchase and operate, ergo, hardly optimizing service costs or contract rates. Of course, the service concept employed by Tulsa frequently involved loads beyond those which smaller vehicles could have accommodated.