A transit agency staffer emails:
I always keep up to date with your blog, and I was wondering if you have any information on revenue/cost ratio calculations on an individual route basis?I am hoping to conduct revenue/cost calculations on individual routes at [our agency], however we have never embarked on such an exercise on a route by route basis, and I have a general idea of how such calculations are done. But I still have some lingering questions.Also, what is your opinion on such calculations? Do you feel they are a helpful tool? Coming from [City x], I have had them drilled into me from when I first got interested in transit, as cost recovery is a big topic [there]. But I notice other areas don't seem to be as interested in it.I was hoping it would be a good tool to show which routes have high recovery ratios and therefore may not only a small amount if any government funding for improved services.
In interconnected urban networks I strongly recommend against emphasizing fare revenue by route, as it creates the illusion that the revenue of each route is an independent result of that route's service. In other words, it conceals the crucial network effect — how routes achieve their outcomes only by working together. If you have financial managers who don't have transit in their bones, they can easily fall into the illusion that the routes are like independent products — different cans on a grocery shelf — and this can lead to some poor decisions.
If the managers really understand interconnectedness in transport networks, then they may find the info useful, but I am still reluctant to prepare data outputs that are likely to be published without their explanations, and this one can be very misleading if you dont' bring that element along.
Costs can also be interdependent if you have a lot of through-routing or complicated operational interlining.
If costs feel reasonably interdependent by route, cost/rider is a better metric to focus people on. This metric values all riders equally, whereas fare revenue by route undervalues transferring passengers and therefore undervalues the interdependence on which great transit networks and their cities thrive.
Transit Agency Staffer,
It’s surprising that any agency involved in allocating taxpayers’ money is not interested in all aspects of operational efficiency and cost recovery, including the performance of individual routes.
This blog doesn’t help, in that discussions are often at a high level and frequently avoid the cost implications of the idealised network structures advocated here. For example, frequent grid networks have lots of service redundancy with a high potential for the misallocation of precious resources. That tends not to get discussed here.
Effective networks are made up of strong, efficient routes/links and in this regard, route by route analysis is essential. Agency staff who know their networks will have a good understanding of what is driving revenues on individual routes, including the proportions of transferring passengers, and how these routes might be made more effective to strenghten the overall network.
Overall measures of network performance going out to the public are great, but people responsible for day to day management need that detailed in-house route level analysis in seeking funding and working towards the overall network outcomes.
“For example, frequent grid networks have lots of service redundancy with a high potential for the misallocation of precious resources.”
No, a grid network avoids redundancy.
Jarett was making the point that you can’t talk about revenue on one route because so many riders take multiple routes as part of their journey.
First, we can start with costs. Any organized agency will have some measure of direct operational costs according to vehicle, way, labor used to offer a given service. It might become a little tricky when we have to deal with ROW maintenance, but nevertheless it should be decompose direct costs as as a function of:
– miles-vehicle driven
– hours in service
– ROW length
– passenger load
Given a proper allocation of direct costs, we need a decent origin-and-destination to open the blackbox of “network effects”. Nothing a good, well-designed survey couldn’t solve – and obviously point-to-point specific ticketing systems such as Oyster card make that information extremely easy to obtain.
One could then perform a step-wise analysis to measure how much each service (a given route on a given frequency served by a given set of vehicles-labor) costs to be compared to the route revenue.
That way, one could capture and even calculate the increase on indirect costs that are incurred since the ridership associated with a certain service is removed from the system.
The tricky part is to use some modelling to predict a “what-if” scenario study when one route is removed in terms of having the ridership completely removed from the system or shifted to overlapping/redundant/competing routes. Not easy, but still doable.
@Nia Maestas: network effects and product interdependence are not something unique to transit.
I read that part of Jarret’s text as “I don’t like when people who are not passionate about transit get hold of its financial management”, a complain that echoes on many activists/advocates in many sectors, where some “uniqueness” is claimed as reason to ignore accounting and cost-analysis practices the rest of the economy is subject to.
This is exactly the mistake that British Railways made during the notorious “Beeching Axe” period in the 1960s: they closed down huge numbers of branch lines, and minor stations on main lines, because they had small numbers of users and were losing money on an individual basis. Then they found that general rail patronage declined far more than expected because users were switching to private cars for their whole journey instead of driving to the closest big station as had been assumed.
When I was working at Houston Metro, they used a complicated route productivity model that included a specific development of fare revenue by route. At the time I was there, they had passes that ranged from annual passes to day passes and everything in between as well as transfers. While the calculations are doable, it is very time consuming especially for an agency the size of Houston. The model included every route that operated weekday, Saturday and Sunday services so it resulted in over 200 different calculations. While the detailed calculations may be useful, it would probably be just as accurate to apply an average fare and you would probably come up with the same results for route evaluation purposes.
I used to calculate fare revenue by route for our agency. It was very tedious and time consuming to calculate with accuracy, but the information was not useful for decision making.
What you end up with is demographic data by route (adults paying full fare vs. transfer passengers, students, seniors, or other half-fare customers). Routes with more of the latter are then penalized for having a lower fare recovery, even though they may haul equal numbers of people.
It is much more valuble (and equitable) to calculate ridership per hour of service by route, which provides a good indicator of how many people board each vehicle for every hour it’s on the street. Additionally, since operational expenses are calculated by hour, this gives a good snapshot of how productive a route is based on what it is costing the agency (Our finance department likes seeing these dollar comparisons, which can be calculated based on our overall direct/indirect cost per hour of service). We’ve switched to this method, and while you still have to be aware of the fact that routes operate as a network, it has become a very useful decision making tool.
I don’t think this particular post is arguing that transit is unique to other products/services. Quite the opposite. As I read it, the point is that transit lines in a well-connected system should be considered complementary products (to expand the grocery store analogy: milk and cereal, peanut butter and jelly, chips and salsa). Together, they generate more demand than the sum of individual lines operating in isolation.
I like to think of this as the Complementary vs Complimentary debate. When private industry offers one product at a discount to generate higher demand for other products (i.e. free salty chips at a restaurant to induce thirst for marked-up drinks), it is the genius of private industry capitalizing on demand for complementary products. When transit systems do the same thing, it is the waste of government offering complimentary products.
@TransitDB: the problem is that when a service relies on direct operational subsidies, the public has at least the right to know the costs.
I remember some transit agency manager once being stubborn in not revealing the costs of paratransit to avoid “stigmatization of its users who have every mobility right as the non-handicapped”, even when no one was demanding it to be dropped.
It appears Jarret Walker adopted a similar position: a “preemptive strike” in not providing data framed in a certain format/concept because people could misuse it as they don’t understand the system (a.k.a, the fiscal hawk of the weekend publishing an op-ed about a money-losing route in the local newspaper citing transit agency data as its source).
I found your challenge to the use of frequent grid networks in transit planning interesting but cryptic. Tell us what you mean by “Effective networks…made up of strong, efficient routes/links” in concrete terms and how this differs from a high-frequency grid.
Anticipating (mistakenly?) that your response will involve “planning routes where and when people want to travel” I feel it is useful for me to share a recent experience of my own.
Having spent a decent part of this past Sunday evening standing on street corners has reminded me of the utility of service frequency. The grid bus network of suburban Toronto made this cross-suburb trip possible, but the waiting time made it undesirable. At least the vehicles were sufficiently full that few would claim the “misallocation of precious resources.”
In this instance the grid was effective at linking origin and destination (with 2 connections in between), in a case where I was the only one to make this given trip, or even follow this route. My line of travel was direct and not meandering. And I was glad to have not needed to plan my entire afternoon around not missing that one bus per day that would supposedly collect all the riders seeking to follow the route I wanted to take.
From my account here you can probably see that I am a fan of at least a particular grid (but just want it to be more frequent) and my analysis is anything but high level. Please explain your proposed alternative.
Sounds a little like a line in my neighborhood which serves a ridegeline during commute periods. Someone wrote to the local paper complaining of the money wasted running an empty bus all day. So I wrote another letter noting the purpose of the line was to bring high school students from the ridgeline to the school on the plain and that there was only one trip for students to take to get to school in the morning and three trips to return them home, then complaining stopped.
Toronto Inner Suburbs,
Network redundancy is a really great thing to have – if your resources are being well used – ie all lines are carrying loads of people, all the time, end to end. This may be the case in Toronto, but it’s not necessarily so in other cities. Depending on the geography and landuse charactertistics of a city (which could possibly be outside North America) other transit network structures may be more applicable – eg radial, hub and spoke, and trunk and feeder networks. I don’t agree that grid transit networks are always the best under all circumstances.
Toronto’s specific, regular grid model is not applicable to my home city (Sydney, Australia) – the geography, landuse patterns and street networks are different – and there are a lot of other factors that might impact on the viability of a Toronto-style grid in Sydney. Part of what makes a strong and effective bus line here is access to and between major employment/retail centres in more of a hub and spoke type arrangement across the metro area. High performance and cost recovery on routes drives further service level improvements. That’s one potential alternative, and you would probably question the applicability of that approach to Toronto.
My main point is that different types of cities need different network approaches, and analysis of individual routes can assist with optimising networks under the varying circumstances. Hope this clarifies a little.
Here in Toronto there is a long history of dubious accounting to establish the “profitability” of services and to save politicians from the difficult task of making policy decision. Reduce everything to a simple number and let staff do the rest.
When this started out, there was an attempt to calculate revenue for each route. This is a difficult and meaningless value in a flat fare, free transfer system where longer trips actually contribute declining amounts of revenue to each trip segment. The original formula actually allocated a fixed revenue per “boarding” with the result that trips taking more than the average number of vehicles allocated more revenue under the model than the rider actually paid. Various fixes were attempted, but eventually the TTC dropped this practice and now only reports allocated costs, mileage and ridership numbers.
Cost allocations have built-in biases triggered by route speed (fast suburban routes consume fewer operator hours to deliver the same level of service as slow downtown operations, while those downtown routes require more vehicles and their associated costs). Choices in the allocation of costs as fixed or variable depending on mileage, fleet size, etc can also be affected by the degree to which a system is dominated by peak-only operation or has a considerable amount of off-peak service and demand. Adding off-peak service can be done at a lower marginal cost (including better labour utilization), but models don’t always pick this up if they don’t distinguish between fully allocated and marginal pricing.
A formula from city “a” won’t work in city “b”.
“Optimising” routes may keep the bean counters happy, but it does not necessarily bring “good” transit service. That word is a judgement and policy call. If, as a fellow commenter from Toronto notes, a well-connected network includes long waits at transfer points, the value of the rider’s time and the disincentive to travel are often omitted from the equation.
There are a lot of transit metrics that are better used at the aggregate level than on a line-by-line basis. Cost per rider and farebox recovery are two examples. Agencies should have overall efficiency goals, then work to find the best solutions to reach those goals. This could involve tweaks to lots of routes, even successful ones, to make them operate more efficiently.
The only metric that I would advocate using line-by-line is riders per vehicle-hour, because it can be accurately measured and really does tell you how popular that particular link in the network is.
Thank you for your clarification. The point of different geographies being more conducive to different styles of networks (and your naming other options) is well-taken.
Although I will not claim a mastery of vocabulary in this field, I read the term “service redundancy” in your initial post as different from “network redundancy” in your follow-up. I conceptualize the former as the presence of vehicles running at low capacity; something I understand to be an inevitability almost every planned system.
Living in a city where our highest elected official endorses a doctrine of “finding efficiencies” in public transit operations, while denouncing any legitimate form revenue generation, I understood your posting to be in the same vein.
That said, I am open to well-reasoned explanations of how systems can become more efficient. For better or worse, however, I tend to be rather incredulous towards promises of greater efficiency, being surrounded as I am by empty mantras of raising revenue by “getting the private sector involved” as I wait longer to try and push my way onto ever busier buses, subways and streetcars.
@Andre, “fare revenue by route” can be a very poor indicator of a route’s usefulness if the system offers free transfer. For example, a lot of bus riders in Queens, New York transfer from subway stations on their daily commutes. Because this transfer is free, those rides wouldn’t be counted, even if those riders were overcrowding the buses. Sure, things other than buses may have network effects and the like, but it usually does not affect revenue to the same degree – I can’t think of any other examples where it essentially equates to “buy one, get another ride free.”
And to add my two cents, the best metric (in my opinion) would be riders per mile. It would heavily slant against express bus routes and the like, but generally, those are also the most expensive to operate per rider (at least in New York.)
One must be careful with metrics like “riders per mile” to define terms.
For example, a route that has a strong line-haul function (the typical thing one finds in express buses) may have a full load, but it is the same set of riders from one end of the trip to the other.
Another route with a lot of local ons and offs may have comparable vehicle occupancy (and the additional problem of friction between riders in the crowd). If one counts “riders” as “boardings”, the local route will outperform the express route even though both have well-used vehicles.
From an “efficiency” point of view, more resources are expended per rider on the long hauls. An express route may be even more costly on a per rider basis because the reverse trip typically runs empty rather than serving local counter-peak demand. This may or may not be an ideal way to provide service depending on the route.
Routes that have many short rides on them (inevitable on the shortest routes) also have low costs per ride and (if one tries to allocate fares) highest revenue per mile. When the TTC tried to do this, their short routes always were “profitable”, but this didn’t stop them from cutting service to make them more “efficient” by filling up unused space. A half-full bus is very “profitable” if it gets a completely new customer load every few kilometres and receives “revenue” based on boarding counts.
The fundamental question about any transit service is this: are people using it? Once you answer that question, then you can turn to policy issues such as fare structure, minimum headways, transfer arrangements. It is the policy decisions that affect the financial performance. Service planning should look first at actual demand and how, if anything, this can be improved.
It’s always amazing how “business practices” are invoked by people who think transit wastes money, but nobody looks at transit like a supermarket. There is inventory sitting on shelves costing money and space, but it’s there because without it people won’t bother to shop. An empty window attracts no customers.
One could easily construct an economic “model” where the cost of building, operating and stocking a store was considered “wasteful”, and say “let them shop online”. What is lost, of course, is the value of getting your supper when you want it, not next week when it shows up courtesy of UPS (or worse after trekking to a local pickup depot).
Economic models tell us a lot about the agenda of those who commission and build them, but not necessarily about the system they allegedly measure. Understanding how a model interacts with real-world customer behaviour and the mechanics of operating a transit system are essential to using them well. Far too many managers and politicians just want a “number”, a “KPI” that will absolve them of actually thinking about policy or accepting the results of their simplistic approach.
Cost recovery is a big deal and keeping the bean counters and politicians happy is an important part of managing a transit system and maintaining ongoing public trust and political support.
I think this discussion highlights the importance of having proper context for any analyses we perform in any industry, not a definitive answer to whether route by route cost to revenue analysis is appropriate for transit. In most other industries, companies do perform “branch by branch” analyses and put findings to good use in many cases. I would argue for transit, we should look at similar context to other industries for such analysis. The context I would most like to see for our system is customer satisfaction. And, of course, we must be aware customer satisfaction, employee satisfaction and profitability (operating ratio in the transit world) move in tandem over the long run. If we really want to fault the route planning for poor performing routes, we should be able to identify the market conditions that were insufficient to support the planned route to confirm planning is at fault. So if your service evaluation is done for any real purpose, it should be to help you identify where “management attention” is needed to bringing some branches up to standard for the organization. Having “cut the service” as the first response to lack luster performance of an individual route is a self-fulfilling prophecy for failing service.