Yes, friends, I am finally starting to dig through the backlog of issues, and I’m starting with an analysis of the TTC’s route cost and revenue figures. These are normally reported in the annual Service Plan, but since there was no plan published this year, the stats for 2005 appear in a stand-alone document at this link.
This data, reformatted as a spreadsheet and with additional columns can be obtained here: 2005 Route Statistics.
I have written before about how untrustworthy these numbers are as a guide to operating costs and service productivity. Various comments came in, and I have held them awaiting a chance to work on this in more detail.
Some time ago, you may have seen the article in the Globe’s Toronto section by Geoff Nixon about the Forest Hill bus and the suggestion by some [misguided] members of Council that routes like this should be axed to save money. I spoke with Mr. Nixon about this article and am quoted in it.
This brings me to the ongoing saga of measuring a transit service for “productivity” and “profitability”.
At the outset, I will attempt to mollify some of the professional planners who say “a bad route’s a bad route, no matter what”. Yes, if a route isn’t carrying anyone, no measurement scheme is going to make it look good. Our problem here is that the TTC’s methodology and arbitrary views about what “productivity” means can threaten routes.
Let’s look first at “profitability”. It is impossible in a flat fare system to make a reasonable allocation of fare revenue to individual routes. The TTC’s scheme goes roughly like this:
- Each time someone boards a new route (the subway counts as one), this is a “boarding”.
- The total fare revenue is allocated to all of the boardings, and then parcelled out to each route pro rata.
- In practice, the average TTC trip involves two routes (some more, some less), and so the revenue allocated per boarding is about half an average fare.
- Trips that are entirely contained on one route would allocate only one boarding’s worth of revenue, and this would understate the revenue due to local trips that don’t need a transfer. The TTC adjusts the revenue per boarding on each route to correct for this although the underlying assumptions in the calculations are not published.
- Trips that use more than two routes (eg bus to subway to bus) allocate more than a full fare’s worth of revenue to their various component routes. This overstates the revenue on routes that serve long trips with multiple transfers.
A major problem with this scheme is that it makes no distinction of how far someone rides on a route, and routes with many short trips (either because the route is short, or because there is a lot of turnover of riders) run up much higher revenues than routes that handle long-distance trips.
On the cost side, the TTC breaks costs down into various factors related to the controlling variable. For example, Operator wages are a function (more or less) of vehicle hours while fuel costs are charged to the distance a vehicle travels. Both of these are subject to distortions, notably the effect on fuel efficiency depending on operating speed and traffic conditions.
The most telling problem with the TTC figures is that although the overall cost recovery ratio for the system is supposed to be 80 percent, the surface system as a whole shows a calculated recovery of only 49 percent. This means that somewhere we have a much more profitable operation — the subway and/or weekend service — or that allocated revenues and costs are out of whack. One way or another the degree of “unprofitability” of TTC routes is very substantially overstated.
Now let’s look at “productivity”.
The most common complaint we hear from people who think the transit system has room to cut spending is that they see scads of nearly empty buses and streetcars. Indeed, the jumping off point for the story about the Forest Hill bus was that it had just one rider.
If we only look at vehicle occupancy as a measure of “productivity” we miss several key points.
- No transit vehicle will ever be completely full all day long over its entire route. For example, the Queen car does not pick up a full load at Neville Park, carry them to Yonge, exchange some of them for an equal number of new riders, and carry all of them out to Long Branch.
- Productivity is at least partly a question of how many people we serve, how far we carry them, what resources we used to do it, and how attractive the service is to existing and potential riders. We may carry a few riders a short distance at low cost, or a lot of riders a long way, and we may lose much more on the full bus than on the half-empty one. Which is more productive?
- Service quality is measured by accessability, frequency, reliability and comfort. If routes are far apart, infrequent, unreliable and jammed with passengers, we are not running a system that will attract new riders and support a transit oriented city. The system is unproductive because it does not achieve the goals we expect of it.
If we compare three routes, Forest Hill, Don Mills and Dufferin, several anomalies show up:
- The variation in the revenue allocated to each passenger. The Don Mills bus appears to have a higher proportion of captive riders than Dufferin. This implies much more local traffic on that route than on Dufferin.
- The much higher daily mileage on Don Mills compared with Dufferin (other factors being roughly equal) does not produce much of an increase in the cost per vehicle. Don Mills runs at an average speed about 25% higher than Dufferin, and this is probably the major factor.
If we look at how many passengers each bus served over the course of the day, we get the following:
- 740 per bus on Forest Hill
- 1,229 per bus on Don Mills
- 1,363 per bus on Dufferin
Many bus routes operate in the range of a 40% cost recovery and under 1,000 customers per vehicle. If we eliminated all of them from the map, we would lose a good chunk of the transit system. Indeed, few routes have the kind of productivity, measured on this basis, that we see on Dufferin. The winner, hands down, is Coxwell with 2367 riders per peak vehicle. This shows very good off-peak ridership plus very high turnover of riders (e.g. very short trips for each rider). Coxwell is an anomaly within the entire bus network and not representative of that mode as a whole.
(The streetcar system as a whole averages about 40% more boardings per vehicle than the bus system due both to vehicle size and the higher turnover of riders in the dense inner city. The winner for the system is Spadina/Harbourfront at 2,411 boardings/vehicle.)
Until recently, we rarely saw efforts by the TTC to increase service on its most “productive” routes by any measure, only moves to cut service. The Ridership Growth Strategy aims to correct this but we have no spare vehicles with which to improve peak service.
The calculation of a route’s cost is alleged to work like this:
(Vehicle hours * Cost per hour) +
(Vehicle kilometers * Cost per km) +
(AM Peak vehicles * Cost per vehicle)
This is a standard three-variable equation that is subject to linear regression (a technique that works backward from the total values and the known parameters to retrieve the underlying cost values). When I imported the TTC’s data into a spreadsheet and split the bus and streetcar routes into two separate groups, I obtained the following results:
Mode Cost/Hour Cost/Km Cost/Vehicle
Streetcar $86.36 $2.56 $489.52
Bus $66.81 $1.21 $150.98
As a check on these values, you can recalculate the allocated cost for each route and compare this with the TTC’s published value. Some small variations are expected due to rounding in the TTC’s published data. For the most part, these values are within a percent of each other, although a few routes are off by up to five percent indicating some anomalies in the data.
The most striking difference here is the cost per hour assigned to the two modes. This cost should be almost completely determined by the drivers’ wages and benefits, plus the supervisory/management overhead. There is no way to explain the much higher value assigned to streetcars.
The cost per kilometer includes variable maintenance costs plus energy. “Variable maintenance” includes those aspects of maintenance that are a function of the amount a vehicle is used. In the case of streetcars, it also includes the cost of track maintenance that should probably be allocated per route mile.
The cost per vehicle is the fixed cost of each vehicle per day. This includes cleaning and servicing plus the cost of operating garages and carhouses. Although streetcars are larger and therefore fixed costs of carhouses are spread over fewer vehicles for a given size of facility, again I have problems with the magnitude of the assigned difference between streetcars and buses.
The inconsistencies in the TTC’s data need to be explained before we can use these factors for a meaningful comparison of the two modes.
Ridership counts vary considerably in their age depending on whether the TTC has the resources to update their data. Large routes get full, detailed counts infrequently, and some counts are at least two years out of date. During a period of ridership growth, this understates the current level of demand relative to the service provided (which is up to date).
Next is the revenue allocated per rider. Because the TTC is a flat fare system, the total revenue is spread out over “boardings”. Every time someone gets on a route (the subway counts as 1 route), this is a “boarding”. Because most riders transfer, there are far more boardings than fares paid.
The fare revenue is not allocated equally to each boarding, but is adjusted to reflect the proportion of transfering riders on a route. A route where many riders’ trips are contained on the route gets higher revenue per boarding, whereas a route with a large proportion of transfer traffic (on or off) gets a lower revenue.
This shows up on the streetcar routes which tend to have more captive riders (riders whose entire trip is on one route), and the average revenue per boarding for these lines is 81 cents compared with 71 cents for the bus routes. There is considerable variation from route to route.
On the streetcar system, this value ranges from a high of $1.07 for Downtowner/Kingston Road to a low of 67 cents for Spadina/Harbourfront. On the bus network, Avenue Road is in top spot at 94 cents per boarding, while Alness sits at the bottom at 54 cents.
Cost, Revenue and Productivity
Another factor affecting the cost of providing transit service is the average trip length taken by each rider. If the typical rider travels only a kilometer or so, the amount of service they consume is low. However, if most riders make fairly long trips on the same route, then they consume more service and the cost per rider is higher.
This typically shows up on three types of routes:
- very short routes where the average trip length cannot be long
- routes with good turnover of riders (frequent ons-and-offs)
- routes with good bi-directional demand (less dead mileage to be shared among the riders).
The bus routes with the highest number of boardings per kilometer are Main, Coxwell, Christie, Wellesley, Greenwood, Vaughan, Parliament, Sherbourne, Dufferin, Dawes, Pape, Junction, Lansdowne. Note that this contains only one long route, Dufferin, that meets the second and third criteria above. The next long route on the list (about five spots along) is Don Mills. The interesting point here is that some of these routes do not have spectacularly good service, but what service there is is well used by this measure even if the buses are rarely full.
The list is similar, but in a different sequence when we look at boardings per hour rather than per kilometer. Since the primary component of service cost is hours, not vehicle kilometers, this would be a better measure. The list here is Main, Coxwell, Dufferin, Royal York South, Christie, Vaughan, Don Mills, Lawrence West, Wellesley, Thorncliffe Park, McCowan, Ossington.
The streetcar and bus systems have very different utilization rates:
Mode Boardings/Hour Boardings/Kilometer
Streetcar 95.09 6.08
Bus 71.41 3.60
Few bus routes have values higher than the average of the streetcar system. This partly reflects the higher demand levels on the streetcar lines but also the larger vehicles and the TTC’s tendency to overcommit their capacity to a higher degree.
Without question, routes that have a lot of boardings relative to the amount of service will always do well in this revenue allocation exercise. Whether the figures are meaningful, especially when paired with suspect cost allocations, is another matter.
Finally, we need to look at the average speed of the two networks. The bus network as a whole averages 19.86 km/hr while the streetcar network averages 13.99. This is mainly a function of the service territories of the routes. If anything, buses operated on streetcar routes will run slower because they don’t do as well in heavy traffic and at busy stops. Most bus routes are faster than the average for the streetcar system, and interestingly the exceptions tend to be the same collection of routes listed above that have very high boardings/hour values. This shows the effect of stop service time on operating speed.
Astoundingly, the fastest streetcar line (on paper) is Queen. I suspect that this is due to the relatively high speed of the Lake Shore segment that pushes up the average. Down at the bottom are Bathurst and Spadina/Harbourfront. On Spadina, we see the combined effects of stop service time and traffic signal delays, while on Bathurst one big effect is the generous allowance for layover time on a short route.
Overall Costs and Revenue for the Surface Network
The total daily cost and revenue allocated to the surface network give us a recovery rate of only 49 percent, well below the nearly 80 percent level of the system as a whole. The total boardings on surface routes are about 1.43-million. Many riders transfer to the rapid transit system, and half of their allocated revenue goes to that network.
However, without the surface network, the subway would starve for business. We must not ignore the contribution of surface routes both as feeders to the subway system for mobility within non-subway neighbourhoods.
Sadly, the values we do not have in the TTC data are measures of crowding and service regularity. We have no way of knowing how many people were packed in like sardines or gave up waiting. Some members of Council (and even some senior management who should know better) take this collection of figures at face value, and make arbitrary proposals for cutting services. This is the “there’s always fat to be trimmed” school of transit management.
The very measures we need to assess service as riders see it are missing from debates over transit service quality and funding.
In a future post on this thread, I will turn to the question of appropriate technologies for various routes.