ALTO Used the Rulebook’s Numbers. It Left the Rules Behind.
ALTO’s new economic report takes two key figures from an Ontario appraisal manual and names that manual as its source. The manual attaches conditions to those figures. The report uses the figures and leaves the conditions out.
In August 2026 ALTO published Canada’s Moment: The Economic Opportunity of High-Speed Rail, an 83-page report setting out the economic case for the Toronto–Québec City high-speed rail corridor. Its central figure is $49.5 billion in benefits to travellers and society, set against a construction cost of $60 to $90 billion.
To turn sixty years of future benefits into one number in today’s dollars, the report needs two things: a rate at which to shrink future benefits back to present value, and a price for an hour of a traveller’s time. For both, it names one source — a public appraisal manual published by Metrolinx, the Government of Ontario’s transit agency for the Toronto and Hamilton region.
The Metrolinx manual does not simply publish those two numbers. It publishes them as part of a package. For any project over $500 million, the same manual requires that the numbers be re-tested at different values, that results be reported as a range rather than a single figure, that early-stage construction costs be topped up to correct for known optimism, and that the project’s benefit-to-cost ratio be published. ALTO’s project is roughly a hundred times larger than that threshold.
None of those requirements appears in ALTO’s report. There is no test of the discount rate, no range around the $49.5 billion, no optimism adjustment to the cost, and no benefit-cost ratio. The two numbers were carried across. The conditions attached to them were not.
This brief does not argue that ALTO used the wrong discount rate. The rate it used is a mainstream, defensible choice. The finding is narrower and, we think, harder to answer: the report presents a set of choices as though they were simply facts, and a reader has no way of knowing that anything was chosen at all.
Why this matters
Imagine a builder quotes you a price for an extension. Asked where the figures come from, they name the standard industry pricing guide. That guide does set those rates. It also says that on a job this size the quote must show a high and a low figure rather than a single number, must add a fixed percentage on top because early quotes are almost always too low, and must set the total against the value of what you are getting.
The builder uses the guide’s rates and does none of the rest. The quote may well be sound. You have no way of telling — and nothing on the page tells you that anything was left out.
That is the situation this brief describes. ALTO’s economic report takes two figures from a public appraisal manual and names that manual as its source. The same manual attaches a set of mandatory checks to those figures for projects of this size. The figures were used. The checks are absent, and their absence is not disclosed.
It matters because of what rests on the result. The $49.5 billion benefit figure is the number now appearing in news coverage and public statements as the reason to build a railway costing $60 to $90 billion of public money, ahead of a federal decision in 2029. Presented as a single figure with no range, it reads as something measured. The report’s own tables call it an upper estimate.
What the report gives you
One benefit figure, $49.5 billion, built on a rate and an hourly value of time presented without explanation of where they came from or what else was possible. No range. No benefit-to-cost ratio.
What its own cited source requires
The same figures re-tested at different values, results published as a range with a confidence level, a 64 per cent top-up on early-stage construction costs, and the benefit-to-cost ratio reported.
What that leaves a reader with
No way to judge how firm the headline number is — and no indication in the document that this is a question worth asking.
One thing this brief does not do: argue that the numbers ALTO chose are wrong, or that high-speed rail is a bad idea. The rate it used is a mainstream choice, and the one test the manual requires would, if anything, make the benefits look larger. The ask is simply that the tests be run and published, as the cited manual says they must be.
What the report borrowed, and from where
The discount rate. A benefit that arrives in 2085 is not worth as much to us today as the same benefit next year. Economists handle this by shrinking future amounts back to a present-day value at a fixed annual percentage — the discount rate. ALTO uses 3.5 per cent a year, applied over a sixty-year period. The rate matters enormously: over sixty years, small changes to it move the headline benefit figure by billions.
The value of time. Most of the $49.5 billion is not cash. It is hours — time that travellers would otherwise have spent on the road or waiting at an airport. To put a dollar figure on those hours, you have to decide what an hour is worth. ALTO uses $22.32, and applies the same figure to every trip: business or holiday, commuter or tourist. The report tells us its own ridership model did separate business from non-business travel, and that this distinction was set aside in favour of one blended figure.
Where both come from. The footnote attached to the discount rate cites one document and one only: the Business Case Manual Volume 2: Guidance, published by Metrolinx in August 2021. The single blended value of time is defended on the grounds that it follows Metrolinx method. So a manual written for regional transit projects in the Toronto and Hamilton area is the published authority for how a national intercity railway has been appraised.
Metrolinx guidance is not binding on a federal Crown corporation, and nothing here suggests otherwise. But a citation carries the terms of the thing cited. If you name a manual as your authority, it is fair to look at what else that manual says on the same page.
The numbers come as a set, not a menu
The 3.5 per cent rate appears in a table of standard parameters. The text introducing that table is direct about their status: any departure from them has to be explicitly agreed during the work, with a clear justification recorded. It is a list of defaults you may leave, provided you say so. Below is how each of those defaults is treated in ALTO’s report.
| What the Metrolinx manual specifies | What ALTO’s report does |
|---|---|
| Discount rate: 3.5 per cent. The rate at which future benefits are shrunk to present value. | 3.5 per cent. Adopted exactly as specified, and correctly footnoted to the manual. |
| Status:Carried across | |
| Value of time: one blended figure. A single hourly value across all modes and all trip purposes, rather than separate values for business and leisure travel. | One blended figure, $22.32. Adopted, and expressly justified by reference to the manual — in preference to the business and non-business split that ALTO’s own ridership model had already produced. |
| Status:Carried across | |
| Growth cap: stop escalating benefits 30 years out. The manual caps growth in the inputs thirty years after the base year, expressly to reflect the fact that nobody can see that far ahead. User benefits are named as covered by the cap. | Applied to two small lines, not the big one. Accident rates are capped after twenty years, and vehicle emission factors run to 2050. No cap is stated on travel-time benefits — which are 78 per cent of the total. |
| Status:Applied selectively | |
| One price year throughout. All values discounted and escalated to a single common year, fixed at the start of the study, so that every figure in the document is in the same money. | Two price years in one report. The appraisal and the capital cost are in 2024 dollars; the economy-wide GDP result is in 2019 dollars. |
| Status:Not consistent | |
| Test the value of time at 0.75 per cent real growth. The manual’s base case assumes the value of an hour does not rise in real terms — but pairs that assumption with a required test of what happens if it does. | Zero growth assumed; no test run. The assumption was carried across. The test that the manual attaches to the assumption was not. |
| Status:Left behind | |
One footnote on the money. Adjusting the manual’s 2021 value of time for inflation to 2024 gives roughly $21.40 — within a few per cent of ALTO’s $22.32. We are not claiming ALTO derived its figure that way, and ALTO does not say how it did. The point is simply that the figure sits where you would expect a Metrolinx-derived figure to sit, which makes the omission of the accompanying test harder to explain as an oversight.
What a project this size is supposed to publish
The Metrolinx manual scales its requirements to the size of the project. Anything above $500 million is treated as large scale, and a specific list of tests and disclosures becomes mandatory. ALTO’s cost estimate is $60 to $90 billion — roughly a hundred times that threshold. Here is that list, and where each item stands in ALTO’s report.
Re-run the numbers at a different discount rate
The manual requires the calculation be repeated at 2.5 per cent so the reader can see how sensitive the answer is to the rate. Not done, and not mentioned. Note the direction here: the required test is at a lower rate, which would make the benefits look larger. Nothing in the manual supports an argument that ALTO’s rate is too generous. What it supports is the narrower point that the manual’s author expects the rate to be tested and the test to be shown.
Re-run the numbers with a rising value of time
Required at 0.75 per cent real growth per year. Not done.
Run the costs and assumptions through a range analysis
The manual requires costs and modelling assumptions be run thousands of times with the inputs varied, and the result reported as a range with a stated confidence level. Not done.
Report the odds that the project is worth doing
The manual asks specifically for the probability that benefits exceed costs. Not reported — the report publishes no benefit-cost ratio at all.
Publish the standard indicators
Net present value, benefit-cost ratio, capital utilisation, return on investment, internal rate of return. None published.
Model low, medium and high growth scenarios
Required, with the ridership growth rate stated for each. A ridership range is shown; the underlying growth scenarios are not stated.
This is not a theoretical requirement
The manual works the method through on a real example: an extension of the Yonge subway line in Toronto, at an early stage of design. A single cost estimate of $5.65 billion becomes a range of $6.16 to $6.84 billion, with the confidence level attached. The reader is shown a central figure, a spread, and how sure anyone is about it.
That project is roughly one per cent the size of ALTO. On uncertainty, the manual ALTO cites tells its reader more about a subway extension than ALTO’s report tells its reader about a national railway.
The report concedes a range it never publishes
The two tables carrying the entire $49.5 billion case are both headed with the words upper estimate. Every figure in the narrative is prefixed the same way: up to 24 million riders, up to 9.3 billion hours saved, up to 400 fatalities avoided, up to 39.1 million tonnes of emissions.
An upper estimate is one end of a range. The other end does not appear anywhere in the document.
The claim built on top of those figures goes further still. The report states that the benefits hold across a wide range of scenarios, and repeats the point in its conclusion. But demonstrating that a result holds across a range of scenarios is exactly what the missing sensitivity analysis does, and exactly what the cited manual requires be reported at this scale. The scenarios may well have been run. Their results are not shown.
And the report clearly knows how to show them. Elsewhere in the same document, the ridership forecast comes with multiple scenarios and a published band around it. The economy-wide GDP figure rests on nearly a hundred separate model runs with the assumptions varied. Between a tested input and a tested output sits the largest single number in the report, presented as a single column of point estimates.
A top-up the manual requires, and the report does not mention
Early cost estimates for big infrastructure projects are, as a matter of record, too low. Not occasionally — routinely. The Metrolinx manual is explicit about the evidence behind this: in an international sample of 258 rail projects, ninety per cent were undercosted, by an average of forty-five per cent.
The manual’s response is a mandatory top-up applied to the construction cost when it is compared with benefits, over and above whatever contingency is already in the estimate. The size of the top-up depends on how far the design has progressed. At the earliest stage — nought to ten per cent designed — it is 64 per cent.
ALTO’s cost estimate is described in its own report as an AACE Class 5 estimate, which is the earliest and least developed class there is. Applied as the manual directs, a cost of $60 to $90 billion would enter the comparison at roughly $98 to $148 billion, before any comparison with benefits is attempted. That range is our own arithmetic on ALTO’s published estimate at the manual’s stated uplift; ALTO publishes no uplifted figure.
The report applies no such adjustment and does not mention the concept. It is worth noting where this reasoning comes from: the manual grounds the adjustment in the research on transport megaproject cost overruns that this Initiative has drawn on since its first publication. That reasoning is already embedded in the appraisal manual ALTO chose to cite.
Too early to divide, but not too early to multiply
The report declines to publish a benefit-cost ratio — benefits divided by costs, the single number a reader would most want. Its stated reason is that the cost estimate is too immature to support one.
Under the framework ALTO cites, that reasoning runs backwards. The manual sets out what is required at each stage of a project’s life. At the earliest stage, the very stage ALTO is at, the requirement is a single line: conduct sensitivity testing to understand the key drivers and the level of uncertainty in each option.
Early-stage uncertainty is not an exemption from testing. It is the reason testing is required. The report treats it the other way round: immaturity on the cost side is given as grounds for publishing nothing, while single-point figures are published on the benefit side of the same ledger. The same uncertainty is treated as decisive for one number and immaterial for the other.
And a business case for this corridor has already done it. In December 2021 the Joint Project Office — VIA Rail and the Canada Infrastructure Bank — published a benefit-cost ratio for High Frequency Rail, the cheaper predecessor to ALTO, at a comparable stage of development: about 0.13, or roughly 0.4 on an expanded basis counting fare revenue and agglomeration as benefits. It published a net present value of minus $21.1 billion and a thirty-year public subsidy of $37.1 to $42.2 billion alongside it. Its economic parameters were drawn from Metrolinx and Ministère des Transports du Québec guidance — the same two sources ALTO cites. Immaturity did not prevent a ratio then. The companion explainer Hours Are Not Dollars sets out those figures in full, including why 0.13 rather than 0.4 is the anchor.
A related point arises elsewhere in the report. In explaining why one set of results is excluded from the welfare account, it refers to those results as therefore not being included in the benefit-cost ratio — speaking of it as a thing with a settled boundary about what enters it. One page says a meaningful ratio cannot yet be produced. Another treats the ratio as already drawn up. The two are difficult to read together, and the report does not reconcile them.
What this brief does not say
Stated here rather than left for others to find.
On the analysis
On the report and its source
Two questions ALTO can answer without releasing a model
Both are answerable from work ALTO has already done. Neither requires disclosure of a model, cooperation from staff, or agreement about what the correct discount rate for a national railway ought to be.
1. Was the calculation ever run at a rate other than 3.5 per cent?
And if so, what were the results? A negative answer is itself informative — it would mean the required test was never performed. An affirmative answer is the sensitivity table the report does not contain.
2. Which parts of the cited guidance were applied, and which were departed from?
The Metrolinx document requires that any variation from its parameters be agreed and clearly justified. The report records no variations at all — while, on the evidence above, departing from several.
Summary ledger
Measured against the requirements of the manual ALTO names as its authority:
ALTO names an appraisal manual twice — once for its discount rate, once to justify a single blended value of time — and leaves behind the testing, the ranges, the optimism adjustment and the benefit-cost reporting that the same manual attaches to those figures at this project’s scale. What remains is a column of numbers labelled an upper estimate whose lower estimate is never shown, resting on parameters presented as facts rather than as selections, in a report that declines to divide that column by the cost while describing the benefits as holding across a wide range of scenarios.
Every document relied on here is public. Nothing in this analysis requires access to ALTO’s models, cooperation from its staff, or a view on what the correct discount rate for a national railway ought to be.
If the terms in this brief are unfamiliar — what a discount rate actually does, why a benefit figure is not money, and who ends up paying — the companion explainer Hours Are Not Dollars covers the same ground in plain language, and sets out the 2021 business case figures in full.