This is the third and final installment in my review of the Metrolinx Initial Business Cases for new GO Transit (and SmartTrack) stations.
Part I reviewed stations on the Stouffville and Kitchener corridors.
Part II reviewed stations on the Lake Shore and Barrie corridors.
Updated March 24, 2017 at 11:00 am: Additional information including replies to some of the questions I posed to Metrolinx and a report of GO Transit’s current ridership added at the end of this article.
Regular readers here will not be surprised at my skepticism regarding the methodology found in Metrolinx reports to perform comparative evaluations of projects. Much of the information is presented at a summary level, important details are omitted, and the underlying assumptions for some calculations are dubious. That said, these reports are the documents Metrolinx relies on to justify its decisions. Understanding how their methodology works is an important part of any critique of the outcome.
What these calculations do not consider is the political context where the “value” of a station is more strongly linked to its perceived delivery on a campaign promise, or to give the impression that transit service will substantially improve where the station is located. This is quite different from how the new facilities will actually affect the transit network or improve the lot of transit riders.
Most of the proposed stations actually do not fare well in the evaluations, although that could well be due to pessimistic projections of the effect of added stops on other “upstream” users of affected GO Transit routes. The evaluation process is very sensitive to the ridership estimates, and if the underlying assumptions change, then so do all of the outcomes.
The new stations to be funded by the City of Toronto have approved as a package, and the detailed IBC reports were not available to Council at the time. A smaller set of stations might make more sense from a financial or planning perspective, but they were sold as a package that is unlikely to be broken up unless some projects prove to be unexpectedly difficult or costly.
A fundamental part of any analysis for a transit investment is the effect on ridership. There is no point in spending millions on a project that would only minimally improve transit demand, provide better service for existing riders, or divert trips from road to transit modes. Transit dollars are limited, and they should be spent where they do the most good.
Contrary to the outlook in some political circles, transit spending should not be primarily a stimulus program for the engineering and construction industries. Too often, notably during the recession of the 1990s, providing good transit at reasonable cost can be a distant, secondary consideration. This outlook still appears in some Metrolinx analyses where the economic spin-off value of spending on construction is treated as a “good”, and therefore more expensive options score more highly because they generate secondary and tertiary economic benefits. That the same money could be spent more productively (“bigger bang for the buck”) or split with other projects or portfolios is not part of the equation. In the case of the new station analyses, this element is not present, and the skew it brings is refreshingly absent.
However, the assumptions behind ridership estimates, let alone details by which the numbers can be evaluated, are consistently missing from the reports, including:
- What is the existing demand on GO Transit corridors, and what are the planned and practical capacities in a decade after the RER network and all-day services are in place?
- What capacity for growth will exist, and to what extent would this be consumed by riders attracted to GO by the new stations?
- What is the assumed elasticity of ridership relative to fares and service? Are these relationships linear (e.g. 10% lower fare generates 10% more riders)? What other factors (e.g. the comparative attractiveness of auto use for specific types of trips) will influence a rider’s mode choice?
- What development effects have been included in the demand model, and what percentage of trips to the expected development will be captured by transit? In effect, does a new station “do its job” in keeping or improving transit’s market share where there is population and job growth?
Some of these questions require decisions/assumptions about “alternate futures” where population and job growth may not be identical, or at least not identically distributed. For example, it was planning gospel for decades that job and population growth had shifted to suburban Toronto, if not to the 905, while downtown was moribund.
This assumption underlies the planning behind what we now know as “SmartTrack”, that more transit to the near suburbs of Milton and Markham was needed, and that growth in the core (and hence the need for more transit service) had stopped. This was a biased view of a study with a strong skew to real estate development in the suburbs, but it went unchallenged, buried as it was in the hype of a mayoral election campaign. Similarly, there are hopes for large-scale growth at the Scarborough Town Centre, just as there have been dreams of suburban centres since the 1960s that simply have not materialized.
It is important, therefore, not to say that “this projection is based on the xyz model of future growth version 3.14”, but to explain what that model actually entails. Only then can a reader evaluate whether the projections might be sensitive to a different outcome and ask what happens with a different set of assumptions.
For the new station studies, ridership is reported in comparison with the “do nothing” option where the station does not exist. Any ridership gains that would occur as a course of other network changes are, therefore, filtered out. The remaining effects are:
- A station attracts new riders as an origin of trips because it is located in a dense residential neighbourhood, or can easily be fed by good transit service from density nearby.
- A station attracts new riders as a destination because it is located in or close to new or existing jobs.
- In both cases above, there is a caveat that some “new” GO Transit riders could be attracted from existing services (TTC or other GO stations) by convenience, and they do not represent a net improvement in transit market share or reduction in road traffic.
- A new station on an existing line adds time to existing trips that could deter riders who would otherwise use the service.
- Additional riding could trigger congestion effects on transit that could limit the rise in demand or even deter some existing users sending them back to alternative travel routes or modes.
These are not the only factors affecting a travel decision. Two important elements are the cost of a journey and the practicality or availability of an alternative choice.
- Fares will affect travel choices, especially where there are significant differences between options. This is at the heart of the Metrolinx “Fare Integration” study which seeks to eliminate “fare boundaries”. That study, however, is hamstrung by a desire to preserve GO Transit’s revenue stream and avoid additional subsidy from Queen’s Park. That is a separate subject, but any new fare scheme will affect how riders choose between routes and modes. The IBCs make no attempt to model an alternate tariff’s effects.
- Practicality is an issue at both ends of a potential park-and-ride or all-auto trip. There is a limit to how much parking GO can provide, and the facilities represents both a capital cost and a loss of development opportunities at large station sites. Parking lots can induce considerable local traffic which is a disincentive in its own right both to drivers and to neighbours. Conversely, an auto trip to a destination may face traffic that is more onerous than taking transit, and there might not be parking available at the end of the trip. This is especially true for trips into the core area.
- Availability is an issue for travellers who cannot afford to drive, or to have a 1:1 ratio of cars to commuters in their families.
The elasticity of demand for suburban travel does not necessarily apply to urban trips, nor is the value identical for peak and off peak journeys where considerations listed above will change quite substantially.
None of this is discussed in the IBC reports. Also absent are specifics on the time of day (peak/offpeak) and directionality (peak/counterpeak) of modelled changes in demand. These do not affect raw counts (numbers of riders), but do affect peak capacity as well as showing how well or poorly the expanded GO network is expected to do for trips that are not the classic peak period commutes.
In all but one case (Spadina station on the Barrie corridor), Metrolinx predicts that a new stop will add 1.8 minutes to a trip from the combined effect of slowing to a stop, handling passengers, and accelerating to the original speed. Each station is evaluated independently, and so the cumulative effect of having more stations is not part of the studies. For example, riders bound to Union from Markham would face four new stations (Finch, Lawrence, Gerrard and East Harbour) if all of the SmartTrack proposals go forward. This would add at least 7.2 minutes of travel time, possibly more, if the station spacing is close enough that trains never get up to full speed over the entire journey. This is actually a more substantial change for 905-based trips than used in any of the modelling.
Conversely, if a change adds only one station, then the effect may not be as great relative to other considerations, and the riding loss could be minimal.
The structure of a service plan is important here, too, because some trains might run express thereby avoiding the time penalty for long distance riders. Losing those riders would run counter to GO’s purpose both in the revenue they represent (longer trips pay higher fares), equipment utilization (empty seats run for miles only to serve short inside-416 trips) and congestion (shifting longer trips away from GO translates to more kilometres of auto use). For most of the new stations, the number of new riders does not offset the projected loss of longer trips “upstream” from the stations. The question, however, is how the factors of practicality and availability will interact with demand and with the change in travel time by corridor, not by station. Fewer extra stations, or the availability of express trains, could limit the loss of riders and make the business case better for some new station sites, especially those with built-in demand like East Harbour.
However, if the political imperative is to add all of the “SmartTrack” stations because that’s what the mayor promised, then comparative evaluations are meaningless (indeed they may never be carried out). The details in the IBCs were available to Metrolinx last summer, but their publication was withheld until after Metrolinx demanded that Toronto Council commit to funding construction of the SmartTrack stations. Several of these stations perform poorly because they are located in low density areas, many industrial, around rail corridors and they are unlikely to stimulate a substantial change in land use.
Two notable exceptions are Gerrard and East Harbour, but even at these sites, some redevelopment is likely to occur whether GO stops at them or not, especially considering that they are also likely spots for stations on the Relief Line which could link the Don Mills and Thorncliffe areas through eastern Toronto to the core. Liberty Village has already developed, and additional residences and jobs here are not dependent on a GO station. The three remaining ST stops (Finch, Lawrence, St. Clair) will not generate substantial demand or development. They are at best alternate routes to the core for residents in their catchment areas, and their attractiveness depends on trip speed (including transfer times) outweighing whatever fare penalty might be involved.
For stations generally, Metrolinx uses a fairly large zone as the catchment area (an 800m radius) as this is “standard” in planning circles. Well, that may be, but a lot of plans using this radius deal with stations dropped into a relatively poor transit environment where the zone of attractiveness will be enhanced by the absence of alternatives. This is not true for an urban setting with a well established local transit network, especially where service on competing buses is already more frequent than the trains will provide, once one gets to the station.
In the case of East Harbour, we know (because the report says so) that additional demand created by future development has not been included in the projections. At the very least, the study should have presented a “before” and “after” East Harbour simulation to show its effect, possibly with a 50% and 100% buildout rather than simply mentioning, in passing in the text, that this development would add riders. This is half-baked work and the report’s conclusions are meaningless.
The studies talk of TTC fare integration, but as we know from Metrolinx studies, this has yet to be defined. As sold in the SmartTrack scheme by then-candidate Tory, and as assumed by everyone who heard him, the “TTC” fare meant full, free-transfer integration between the existing local network and “SmartTrack” at current fares. That is not what Metrolinx has discussed in the context of fare integration where a fare-by-distance model would apply at least to rapid transit services (subway and LRT) on the TTC, and these higher fares would be the same as for GO trains.
Further unanswered questions include:
- What is the effect on demand of a “TTC” fare at existing stations in the SmartTrack corridor (Milliken, Agincourt, Kennedy, Main, Union, Bloor, Weston, Etobicoke North)?
- Will the “SmartTrack” fare be applied to trains in other corridors within the City of Toronto (Lake Shore, Richmond Hill, Barrie, Milton)?
- How much of the difference in fare revenue will be expected as a City contribution to GO’s budget?
It is self-evident that as GO Transit morphs into a local service provider with a cheaper inside-Toronto fare that demand will rise, and this will occur on the most congested part of its routes just as the TTC subway is overloaded the closer one gets to the core area. What has not been explored is the location of the “knee in the curve” where the combination of cheaper fares and more frequent service causes a non-linear change in travel shifts to the GO network.
Metrolinx financial evaluations can make frustrating reading because so many components are rolled together into a single metric that “value” can be meaningless. The premise is that various schemes and options are reduced to a single number alleged to represent the long-term benefit or cost ratio. However, evaluating a project is rarely this simple.
Any scheme will entail both capital and operating spending, plus the financing costs of any debt or leaseback strategy. These costs are rarely borne by a single agency or entity, and the situation is further complicated where a municipally sponsored project is cost shared with the province. Benefits flowing from a project do not necessarily flow to the same governments in proportion to their investment or ongoing costs, and moreover the soft benefits (such as imputed time savings for transit riders) generate no revenue.
Although the studies consider projects over a 60-year lifespan, they do not break down the cash flows for elements of the life cycle. This is important for financial planning because no matter how much we may want “everything”, it may not be practical to undertake every project in the same time frame. This consideration affects the City of Toronto’s plans to control the growth of debt and associated servicing costs, and it has affected provincial decisions about the timing and number of projects, notably their on-and-off commitment to various elements of Transit City. Notable by their absence from all of the projections is any provision for capital financing whether this be through direct borrowing, or by third party arrangements such as a PPP where financing costs show up as an ongoing operating charge against a project.
The soft benefits produce “savings” for society in general, but a good deal of this occurs in the latter years of a project’s lifespan, and in any event cannot be recouped by governments to pay down debt or offset day-to-day operating expenses. A project may have a positive “value” on paper, but this is in the wider sense that public investment creates a future public good. However, this investment still has to be paid for.
Secondary costs and benefits such as additional network capacity costs or avoided expenditures through redirection of traffic rarely show up in these evaluations, particularly when individual projects might combine to trigger effects beyond those of any one in isolation. A very large example of this is the provision of “relief” capacity which can make a network more robust and avoid the need for upgrades to other network elements. On the scale of added GO stations within the 416 and encouragement for transit riders to use GO as a local “rapid transit” link, the effects of many changes – numbers and locations of stations, service levels, fare policy – will be different from the sum of individual station projections.
Finally, the use of a long time frame, 60 years, for evaluation encounters two problems. First, it assumes that the short-term change attributed to a project will continue more or less as is over the entire period, and second it brings imputed soft savings from the distant future into calculation of the present worth of a hard investment.
To be fair, those future savings are discounted, but the problem with long term forecasts is that they are sensitive to assumptions about inflation and discount rates. However, the combination of several factors usually masks how this works. There is one example in the station studies where this is easily visible, however. The Park Lawn Station replaces Mimico station, and its only effect on ridership is to add 10.3 million trips with no offset for lost riding elsewhere. The projected additional revenue is $13 million, or roughly $1.30 per trip, well below the minimum GO Transit fare. This is the result of discounting future fare revenue to a present value at a rate higher than the presumed rate of increase in fare levels. This revenue is at least “hard” money in the sense that riders will actually pay to use the trains, as opposed to the “soft” money that will be their imputed future time savings.
Ridership effects are central to these projections because they are related to:
- fare revenue
- the imputed value of time saved or lost in travel
- changes in vehicle kilometres with riders shifting to or away from transit
- the value of auto operating costs
- the imputed value of reduced road congestion
- the imputed value of environmental benefits
Central to Metrolinx work on “The Big Move” has been the concept that transit investment will reduce road congestion. This is a myth that even Metrolinx has long acknowledged. Getting more people on transit will stem the growth in demand for road travel, but it will not reduce congestion. In fact, there will be areas where congestion gets worse because the Metrolinx network is so strongly focussed on travel to downtown Toronto and does not address the suburb-to-suburb travel needs that produce some of the worst congestion today.
Similarly, environmental benefits assume a reduction in pollution and secondary effects such as health benefits. At best, the growth of emissions will be lower than it might have been without the Metrolinx projects.
In many of the station analyses, there is an assumption that an extra station stop will drive away riders, often more than are gained by the new station. Moreover, the lost riders will have been making longer trips, and so factors such as road congestion take a bigger hit from losing these transit riders than the benefits of the new short trips. There is also a good chance that many of the new GO trips were already on the TTC and they do not represent a net gain for transit.
Media and political coverage of these studies emphasized that the new stations meant a net loss for GO, but I am not as convinced as the study’s authors that a small added travel time would actually do this. The effect might be stronger where multiple stations combine to make trips longer especially if the new stations are added simultaneously after riders are used to faster trips. Then there is the question of the offsetting benefit of electric versus diesel hauled trains on travel times, and how a change in motive power could offset the additional stops.
At the very least, the station proposals should be reviewed to determine their potential effect as groups, not as single projects. A better understanding of just how added travel time might drive away riders at the projected level is needed. What will affect their decision? Do they have a practical alternative end-to-end? A simplistic elasticity factor does not reflect these considerations.
On a network basis, understanding of capacity is critical to evaluation of any proposed change. I have asked Metrolinx for current ridership by corridor, and for the projections cited in footnotes in their studies, but have yet to receive them. If new stations reorganize ridership demand and discourage long-haul travel, this is counterproductive. The idea of using the rail corridors as rapid transit lines is a strong one in theory, but only if sufficient capacity will be provided for both long haul and local trips. The potential cost of lost capacity is not included in any of the evaluations.
The following section summarizes the elements included in the calculation of a station’s “value”.
Metrolinx calculates these values over the 60 year life of a station allowing for inflation and with future values discounted back to 2015.
- Incremental ridership gain or loss together with the associated change in fare revenue over the project life, discounted to present value
- Capital costs including major renewals, but not including ongoing financing or leaseback charges
- Operating costs over the project life (discounted)
- Station operation and maintenance
- Incremental train operating costs
- Leasing revenue at stations, if any
- Land value capture
It is unclear whether train costs are adjusted only for fuel/power consumed for an extra stop, or if there is an assumed extra cost for the train crew. The latter should only apply where the extra time needed for stops is sufficient to trigger a change in the number of trains required to provide service.
Land value capture is a concept that has underlain the SmartTrack scheme with the assumption that new stations will generate development, an uplift in property values and more tax revenue that would offset the cost of the ST system. This was a dubious claim when ST was first proposed, and is even more tenuous now that the service level and station count is well below the level in election claims. In any event, there remains the question of how any uplift in property value and tax would be shared between Queen’s Park and Toronto given that both will invest in the stations, line, service level and fare discounts such as they may be.
Various possible sensitivity analyses are listed that could affect the projected costs and revenues, but these receive only superficial discussion, not worked-out examples. This affects scenarios such as alternate station sites, service levels and fare structures. Only a single set of numbers goes forward to the next step.
This group of costs and benefits depends strongly on the projected changes in ridership, travel time and travel mode.
- Time savings in millions of person hours
- This translates to an imputed value of time saved
- Auto distance savings in millions of kilometres travelled
- This translates to hard savings in vehicle operating costs (the primary component of this factor), plus a value for decongestion of the road network, safety benefits of less traffic, and environmental benefits of lower emissions.
In each case, costs and savings are inflated to future years, and then discounted back to present values. These values are weighed against the capital and operating costs listed under the Financial Case above.
Other factors considered include:
- Deliverability of a project – could it actually be built, and what effect would there be on operations either for GO or other agencies
- Strategic fit of a project with factors such as:
- local and regional planning policy,
- potential for development,
- effect on the natural environment,
- effect on the GO network,
- station access, and
- social factors such as the type of neighbourhood served.
These are evaluated on a broad scale of “Supportive”, “Neutal” and “Not Supportive”.
For a full discussion of these factors, refer to the appendices of any of the IBC reports.
Updated March 24, 2017 at 11:00 am:
Questions and answers from Metrolinx:
Q: Do you have current ridership counts (say daily or weekly) by rail corridor? The quick fact sheets only give values for the full system.
A: Attached is a report detailing ridership for September 2016. The document reflects current rail service patterns, excluding weekend Barrie rail service and Richmond Hill service extension to Gormley GO Station.
The corridor and route breakdowns provide interesting reading, but the important issue here is to compare current daily and monthly ridership with the projected gains and losses over the 60 year span of the station studies.
The Stouffville corridor, the 2nd least used of GO’s rail services, carries 333k riders/month. If there were no increases for the next six decades, the total would be 240 million rides, and we know that the numbers will be higher thanks to planned improvements in frequency and hours of service. Ridership changes of 20m over 60 years represent less than 10% of this number, and the proportion will be smaller when future growth is factored in.
While any loss is a subject for concern, it should be stated in context especially if, as I have argued, the change in trip times may not be sufficient to trigger the projected losses.
Q: In the IBCs for Lawrence East and Ellesmere, the effect of adding 1.8 minutes to the running time is shown as a loss of about 49 million rides over the 60 year timeframe of the study. However, the IBC for Finch, which would incur the same penalty, shows a loss of only about 20 million rides. Since most riders on this corridor originate north of Steeles, most of them are affected by the addition of one station no matter where it might be. Why is the effect so much lower at Finch East?
A: Impacts to upstream riders are similar at all three stations. Ellesmere and Lawrence would delay Agincourt GO passengers. Since Finch is north, it would not impose the same delay to Agincourt passengers.
Finch also anticipates higher potential for new boarding, both locally and through bus transfer, which offsets the loss of ridership in the calculation of overall impact.
Tables in the reports clearly show riders lost due to extended travel times separately from new riders attracted to the stations. If the effect from stations at Lawrence or Ellesmere would be so much greater than at Finch, this implies that Agincourt is already a major source of trips and that these trips would be discouraged by the addition of one stop further south. This does not make sense.
Q: Park Lawn station is unusual in that it involves the loss of no existing riders because there is no time penalty in moving the stop from Mimico to Park Lawn, and there is no projected drop in ridership. Over the course of the 60 year life estimate, the change in station locations is projected to generate 10.3m new trips, but only $13.1m in new revenue, or $1.31 per trip. This is lower than any fare GO charges.
A: The Park Lawn analysis was a unique case in the set of IBCs, as it was evaluated as a replacement station, not an additional station. No time penalty for upstream riders was incurred because it was assumed the time impact of Mimico station would simply be transferred.
There would be some loss of existing riders, particularly due to a loss of parking at a Park Lawn station, which is currently a significant source of ridership at Mimico. New ridership at Park Lawn was based in part on assumed higher residential densities near Park Lawn vs the existing Mimico site.
The table speaks to 10.3M new boardings and alightings, not net new trips. Alighting riders at a replacement station would not be considered to add to fare revenue. Also, a discount rate is applied to financial calculations throughout, so “fare revenue” does not directly translate into 60-years of farebox payments.
If only new boardings count as a revenue gain, then this understates the network effect. Everyone who boards as a new inbound rider also becomes a new outbound alighting, and they pay fares both ways even though only one of the fares is paid at Park Lawn. A trip that shifts from Mimico to ParkLawn (regardless of direction) for convenience is clearly not a net new rider.
As Metrolinx states, some riders who now park at Royal York may be lost if the station shifts, but they do not appear as lost “upstream” riders. The problem here appears to be that too simplistic a representation of the various rider classes and changes has been used.
As for the low average value of a new ride, this is likely due to the combined effect of presumed future inflation of fares and discount rates to bring future revenue to 2015 dollars. The discount rate is higher than the inflation rate, and so future revenue generates less money expressed in 2015. This drags down the average value of a fare. Other factors may be at work too due to Metrolinx’ consolidation of multiple ridership effects in a single number.