[This article has been linked from torontoist.com where there is another thread of comments.]
Whenever there is a budget crisis, the TTC trots out its annual report in which they claim to show the costs and revenues associated with each route in the system. By implication, the routes at the bottom of the barrel are “poor performers” and candidates for service cuts if not outright extinction. The calculations in this table can be charitably described as creative writing.
In a flat fare system, it is impossible to allocate fare revenue in any way that makes sense and produces meaningful comparisons between routes.
The methodology used by the TTC goes something like this:
- Every trip (and hence fare paid) represents one or more “unlinked trip” in planning lingo. Every time you change vehicles, that’s a new unlinked trip or “boarding”. It turns out that the average trip consists of roughly two boardings. In other words, the average passenger transfers once. Some don’t transfer at all, some transfer twice or more.
- The average fare is divided by the average number of boardings per trip to produce the average revenue per boarding.
- The subway counts as one route regardless of how many times you transfer. I am not sure about the SRT because none of the rapid transit lines appears in the annual statements (more about that later).
- The TTC cooks the figures to correct for the fact that some routes have a higher proportion of riders whose entire trip is on that route. They should contribute a full fare to the route’s revenue. Of course, this leaves less revenue to spread around among those who transfer. There is no documentation of the underlying assumptions, but the stats clearly show that wildly different values are assigned to the revenue per boarding on different lines.
Cost allocations are actually done on a more reliable basis thanks to work by me and others many years ago to identify underlying problems with “average” operating costs.
- Costs which tend to be sensitive to the number of vehicle hours (notably operator wages and, possibly, some aspects of vehicle operation) are assigned to vehicle hours.
- Costs which tend to be sensitive to the distance the vehicles travel (notably routine maintenance and fuel) are assigned to vehicle miles.
- Costs which tend to be sensitive to fleet size (notably garaging and routine cleaning) are assigned to peak vehicles.
There can be some debate about the fine points of these allocations, but they are generally defensible. However, they are not directly useful for comparisons between routes with vastly different operating characteristics. For example, we all know that fuel consumption goes through the roof in stop-and-start traffic, and you cannot take the average fuel consumption of the bus system and use it directly to project the cost on a congested route like Queen.
Let’s look at the most recent statistics published by the TTC for the 2005-2006 period. Something should jump out at you on the very first line: the 192 Airport Rocket has a cost recovery of only 22% and yet it never shows up on the list of “Poor Performing Routes”. Even more amazing, we know that this line is overcrowded and that TTC has been forced to add service. Why are the statistics so bad?
Simple: It is a line haul route that picks up passengers at Kipling and delivers them to the airport. There’s a good chance that when it’s busy outbound, it’s quiet inbound. The number of boardings per unit of service operated is considerably below that of the bus routes collectively. Regular routes have a lot of turnover with passengers boarding and alighting all along a route. This yields many more boardings (and hence allocated revenue) per hour or kilometre of service, and a better cost recovery number.
Another interesting point about the 192 is that it has the highest cost/hour of service, but the lowest cost/km. This is easy to understand. The average speed of the vehicles is double that of the rest of the bus system and so for each hour of operation, the costs related to mileage components are much higher than on the rest of the network. Conversely, when we look at the cost/km, the high average speed dilutes the cost of the operator and we get a much lower value than on the system as a whole.
By now you are probably wondering whether I have already sliced and diced the TTC’s figures to make life easier for analysis. Here is the same data presented with several additional fields.
This version reveals a lot more about how costs and revenues have been allocated by the TTC. The added fields are:
- Boardings per hour. These give a measure of vehicle utilization in terms of turnover, the number of people who get on (and off) per unit of service. Coxwell 22 is the high roller among on an hourly basis because it is a very short route (Danforth to Queen), it has strong bidirectional loading and it has a good turnover at its midpoint, Gerrard. However, both the Spadina and Dundas cars beat it out on the system overall. These routes handle a lot of short trips.
- Boardings per kilometre. These values behave similarly to the hourly values, but due to differences in the average speed of each route, the relative positions in the list will change. Among the buses, Main ranks highest followed by Wellesley. The values for all of the streetcar routes are higher than the average for the entire bus system. This reflects both the density of demand and the larger size of the vehicles.
- Revenue per boarding. This shows how different values are used depending on the imputed proportion of one-seat rides on each route. Note that the streetcar routes have a higher revenue per boarding (81 cents) than the bus routes (71 cents) reflecting the higher proportion of one-seat trips taken on streetcar routes that serve both the origin and destination for many neighbourhoods.
- Cost per boarding. This value varies a lot between routes. For the streetcar system, the cost per boarding is higher than on the buses due to the much lower average speed on the streetcar routes. However, the ratio between these costs is much lower than the ratio of speeds because of the larger vehicle capacity on streetcar routes.
- Revenues and costs are also shown per hour and per km along with the average speed. Note that this is the actual number of kilometres divided by the actual number of hours in the TTC tables and gives the all-day average including garage mileage. These numbers will be different from those shown in the TTC schedule summaries.
- The values are totalled for each mode and for the system overall.
The real standout in these figures is the system total cost recovery of only 49% for the surface routes. This shows that a huge amount of revenue is allocated to the subway system in the model. More importantly, when people discuss cost recovery, the surface routes are already at a disadvantage because so much revenue goes to the subway system.
The “sorted stats” file contains two versions of the same data, but they are presented differently than before.
The first three pages show the routes ordered by percentage cost recovery, and the routes earmarked for cancellation are highlighted in red. Note how many routes have recovery rates lower than those to be cancelled. The usual explanation for their survival is that cancelling them would inconvenience too many riders, or would cost to much in lost riding, or a variety of other excuses.
The fundamental point is this: when we talk a bout “poor performing routes”, we should be talking about routes that don’t do very well, but too many people will complain if we cut them.
Another notable point is that some of the routes proposed for elimination have cost recoveries very close to the system average of 49% for surface routes: Bloor West, Pharmacy, Mortimer and Lambton are at 46% or better. Meanwhile, routes like York University, Scarborough, Highway 27 Rocket, and Steeles East are below 40%.
The figures published by the TTC cannot be used to identify candidate routes for service cuts. However, the TTC persists in painting these routes as the bottom of the barrel, the ones that just have to go, when in fact many surviving routes, routes where service improvements are planned, do as badly as routes that are on the chopping block.
Pages 4-6 of the sorted statistics show the same data again, but with one column added: the imputed daily loss for each route. For many reasons discussed earlier, this figure is very suspect because it depends on the revenue allocation technique. In some cases, overlapping routes are joined as one for reporting purposes, while in others they are separate. This exposes some routes to scrutiny and hides others.
As I said before, routes that provide only tripper service such as 503 Kingston Road and 508 Lake Shore will always have poor cost recovery as we would expect from peak-0nly trips on any route where loading is unidirectional and the proportion of truly productive mileage operated is low.
The intent of this table is to compare with the stated loss on the Sheppard Subway of $10-million/year. This is actually the net cost of running the subway rather than the bus routes it replaced, and this number is subject to verification. In any event, a back-of-the-envelope rule for converting between weekday and annual costs is to scale up or down, as appropriate, by a factor of 300. This presumes that all of the weekends and statutory holidays are equivalent to about 50 weekdays, and that there are 250 regular weekdays in a year. One could quibble, but the numbers will be correct for the purposes of rough comparison.
This gives us a daily cost of the Sheppard Subway of about $33,300, and you can compare this to the rest of the system. The Finch East and Steeles East buses are in the same league as is the Carlton streetcar. Again, this is subject to the faith that you place in the TTC’s revenue and cost allocation schemes.
Far be it for me to turn into a Sheppard Subway booster, but it’s important that everyone understand that running major routes costs a lot of money even with buses, let alone a subway network.
The TTC is feeding the public and Council a load of baloney when it talks about “Poor Performing Routes” because these have been plucked from a much longer potential list and carefully examined to minimize the fallout from any cuts. Many routes lose far more per rider than those in the dreaded lists, but they are immune.
Before we start running around the transit network axing suspicious routes left and right, it’s important that we remember that this is a network and when you start taking parts away, you damage more than the individual pieces. That’s what happened in 1996, and it will happen again if we don’t fight for the system as a whole.
Tomorrow (or maybe the day after), I will review the proposed cuts to see which Councillors’ wards are most affected. “Before” and “After” route maps for each ward would look nice, but I will leave that to the folks who have a graphics department to pull that together.
Postscript [July 23, 6:40 pm]
A comment on this item from Jonathan prompted me to review recent TTC reports. He asked whether decisions on routes were being made with out-of-date data.
The numbers in the tables here are taken from the 2005/06 reports. More recent data is available, but it has not yet been published. Here is the chart that appeared in the Star on July 15. [If anyone at the Star has copyright issues here, please let me know.]
As you can see, the passenger count has been flat for a few years, and it took a real beating from the combination of the early-90s recession and the mid-90s service cuts. For 06/07, there has been a big jump back to a level we have not seen since 1991/2. The jump in 1997 is caused by the conversion of the Spadina line from bus to streetcar. Some RGS-related improvements came into the streetcar system, but the jump in riding is disproportionate to that.
Therefore, the performance of the streetcar system as a whole is underreported in the tables I have published. There is no comparable published data for the bus system, but I would not be surprised to see strong growth there too.