What our build odds rest on
We followed nearly 400 past Canadian resource projects to see how often each stage ends in service. About three approved projects in ten were running within eight years; under construction, nearly nine in ten. How the count was made, and where it is weak.
Updated since publication: 3 updates
Also in French: Sur quoi reposent nos probabilités de construction
Our forecast's odds that a project gets built rest partly on a count of past Canadian projects. Of the projects that stood at the approved stage on 1 January 2014, 2016 or 2018, about 29% were in service within eight years. Of those under construction, about 88%. Of those postponed, about 4%. Our forecast, What Canada nets and Does opposition stop projects? all lean on that count. Readers couldn't check it until now. Here is how it was made, where it is weak, and its sources. The coded table of projects is on our open data page.
| Stage on 1 January (2014, 2016, 2018) | Projects counted | In service within 8 years | Our judgment before the count | What the forecast uses |
|---|---|---|---|---|
| Proposed | 71 | 15% | 20% | 20% (the count's range includes it) |
| In review | 66 | 12% | no such stage in the forecast | none |
| Approved | 49 | 29% | 50% | draws built on the raw count, which centre a little lower: 20% to 31% in eight runs out of ten |
| Investment decision taken | 12 | too few to read | 85% | 85% |
| Under construction | 66 | 88% | 90% | 90% (the count's range includes it) |
| Postponed | 52 | 4% | 10% | 4%, drawn from its range |
The big drop comes after approval. Approval is not the last hurdle: seven approved projects in ten were not running eight years later. Once building starts, nearly nine in ten are running within eight years. And a postponed project almost never comes back: eight in ten were still on the shelf eight years on.
What we counted
Nearly 400 projects, followed to an outcome. We started from the lists governments keep of major projects: Natural Resources Canada's national inventory, British Columbia's quarterly inventory and Alberta's major projects list, editions from 2005 to 2025.1234 We kept resource and energy projects of C$100 million or more: oil and oil sands, gas processing and LNG, pipelines, mines, power lines and large dams, and export terminals.
Every large project, a sample of the smaller ones. We coded every project of C$1 billion or more and a random sample of the smaller ones, then weighted the sample back to the full list. That gave 379 projects in the main count. On 30 June 2026, 152 were built, 21 under construction, 75 still moving, 76 shelved and 55 dead.
Each stage dated from the record. For every project we dated each stage it reached (announced, filed for review, approved, investment decision, construction, in service) from regulators' and courts' decisions, company filings and government records, including the federal assessment registry,5 not from the inventories' own labels.
How a rate is made. On each 1 January from 2006 to 2018, every second year, we noted each live project's stage, then checked whether it was in service eight years later. Eight years, because the forecast looks to 2034. A project counts again at each date it sat in a stage, so the 49 approved projects give 88 observations. Before 2014 only British Columbia and Alberta are covered, so the forecast uses the national years, 2014 to 2018.
What it did to the forecast
A rule decided when the count could replace our judgment. Written before any project was coded: take the count for a stage when it rests on at least 20 projects and our judgment falls outside the count's likely range. Approved (50%, against a range of about 20% to 38%) and postponed (10%, against 1% to 8%) failed that test, so the count replaced them. Proposed and under construction passed, so our judgment stays. Too few projects sat at the investment-decision stage on those dates to say anything.
Fast-track projects borrow the approved odds. Projects referred to Ottawa's new major projects office or listed as of national interest have no history to count. Under the same rule, their odds must sit between proposed and approved; ours sat above, so they now take the approved value.
Lower odds of building barely moved the forecast. Most of the large projects at those stages today are bound for the coast: Ksi Lisims LNG is approved, the Pacific Link oil pipeline is listed. In our model, each one built sends resources to the coast rather than south, but about a third of new coast oil still reaches the US by tanker, judging by where Trans Mountain's tankers have gone. So making them less likely raises the forecast by less than half a point. A postponed project, such as Enbridge's Mainline Optimization Phase 2, now counts for little.
Event · 2025-09-15 · action · CAKsi Lisims LNG receives its environmental assessment approvals
On 15 September 2025 Ksi Lisims LNG received a federal Decision Statement and a provincial Environmental Assessment Certificate (Major Projects Office). The project would export up to 12 million tonnes a year from up to two floating liquefaction units on BC's north coast, supplied by the roughly 750-kilometre Prince Rupert Gas Transmission pipeline. It was referred to the Major Projects Office on 13 November 2025.
- Major Projects Office, “Ksi Lisims LNG - Major Projects Office”.
- NGI, “Ksi Lisims LNG in advanced discussions for more offtakers as it targets FID by year's end”, 2026. paywalled
Event · 2026-10-01 · action · CACarney lists the West Coast oil pipeline, renamed Pacific Link, as the first project of national interest under Schedule 1 of the Building Canada Act
In Fort McMurray on 1 October 2026, Prime Minister Carney listed the West Coast oil pipeline, now called Pacific Link, as a project of national interest under Schedule 1 of the Building Canada Act, the first project designated this way (CBC). The pipeline would carry about one million barrels a day roughly 1,250 km from Bruderheim, Alberta, to a deepwater port outside Delta, BC (Global News), on a southern route that the government says protects BC's North Coast and the Great Bear Sea. Canada and Alberta are to hold equal shares, Pembina Pipeline is the private investor and Trans Mountain Corporation leads development; Indigenous communities will be offered at least 10% ownership, financed through loan guarantees. The Major Projects Office is to finalize the conditions by 1 September 2027 (Prime Minister's Office). No final investment decision or in-service date was announced.
- Prime Minister's Office, “Prime Minister Carney lists the West Coast Oil Pipeline, now known as Pacific Link, as a project of national interest”, 2026-10-01.
- Canada Gazette, “Order Amending Schedule 1 to the Building Canada Act, P.C. 2026-925 (Canada Gazette, Part I, Vol. 160, No. 41)”, 2026-10-10.
- BOE Report (The Canadian Press), “A timeline for Pacific Link, the national interest Alberta-B.C. pipeline plan”, 2026-10-01.
Event · 2026-07-31 · action · CAEnbridge postpones Mainline Optimization Phase 2 (+250,000 barrels a day, Alberta to North Dakota) because producers won't commit
Enbridge said Canadian oil producers had not committed to the 250,000-barrel-a-day second phase. Coverage cited oil-price volatility since the start of the Iran war and the fact that the federal–Alberta energy agreement remains non-binding. Phase 1 (about 150,000 barrels a day) reached a final investment decision in November 2025.
- Global News, “Canada's Enbridge postpones plans for second phase of Mainline oil pipeline expansion”, 2026-07-31.
- East Daley, “Enbridge Shelves Mainline Optimization Phase 2, Doubles Down on Phase 1”, 2026-08-12.
Where the forecast stands. It puts 96% on clearly more Canadian resources reaching the US in 2034 than in 2024. On the narrower question in What Canada nets, whether the course taken since January 2025 paid for itself, Canada now comes out ahead in only 44% of our runs. Both changes are in the corrections log.
Current odds · outcome
These odds come from our models, whose rules are written down before they run; each row is the result of its model’s current official run. Rows can overlap (an “or higher” row includes the rows above it).
| Code | Explanation | Odds | Range | As of |
|---|---|---|---|---|
| O0 | No meaningful speed-up | no estimate | ||
| O1 | Faster, but the extra output goes mostly elsewhere | no estimate | ||
| O2 | Faster, and more volume reaches the US (this level or higher) | 96% | 2026-11-12 | |
| O3 | Locked-in preferential terms for the US | 26% | 8%–45% | 2026-11-12 |
The uncertainty goes in too. The count is a sample, so the forecast doesn't use 29% as a fixed number: each run draws the approved odds from a range built on the raw count, in eight runs out of ten between 20% and 31%. Why that range sits a little low is the second of three choices set out below.
What the count can't tell
Why projects die, mostly. Of the 55 dead projects, 30 have no record showing a decisive cause, and 5 were not coded for causes. Where records name causes, they are spread out: regulators, governments, financing and markets each appear for between about 11% and 18% of dead projects. Companies most often blamed markets (19 projects), but other records back that up for only 4. Shelved projects are almost never explained: 84% have no record showing a decisive cause.
Opposition shows no overall effect. Across all stages, projects facing any challenge (a court case, a formal intervention at a hearing, a campaign or a government action) reached construction within eight years a little more often than the rest: 46% against 40%. Sustained campaigns were rare, 10 projects, and only about one in ten of those reached construction, against 42% of the others. Overall, challenges by First Nations, Métis or Inuit groups made almost no difference.
One stage shows a large gap. At the approved stage, projects with any challenge reached construction 26% of the time, against 52% without one (29 and 34 projects); challenges by First Nations, Métis or Inuit groups alone show a similar gap there.
Too few cases to enter the forecast. The study's rule for adding opposition to the forecast tests only campaigns, alone or with court cases, and needs 20 such projects in a stage; none has more than 10, so nothing was added. A challenge may follow trouble rather than cause it, and the biggest projects draw the most: among projects of C$1 billion or more, challenged ones reached construction more often.
How well the rules held
Written first, frozen, then applied. The coding rules were frozen on 19 October 2026, before any project was coded. A second coder, working blind, coded a random 68 projects; both are AI research agents working from saved copies of the records, and the second never saw the first's codes. Agreement was scored with a standard measure that discounts chance, against a target of 0.7 (1 is perfect agreement).
Outcomes held; challenges didn't fully. Agreement on outcomes met the target (0.79, then 0.91 once the rules were clarified). Main causes met it after clarification (0.88, from 0.52); stages only at the margin (0.70, from 0.69). On challenges, as first coded, only court cases met the target. After the rulings below, formal interventions and "any challenge" met it too, but campaigns and government actions never did, so the opposition results are descriptive. One slip: a briefing error exposed the blind coder's challenge codes, so the checked projects' challenges were recoded by a fresh coder.
Rulings made during coding are marked as such. Three rounds of clarifications and 94 readings of the rules on particular cases were written after codes had been seen, some after an earlier run of the results. Each rests on the rules' wording, not on its effect, and each that is a judgment also runs the other way. None of those alternatives moves the approved or postponed figure by more than a few tenths of a point.
Other choices matter more. For approved projects, dating stages known only to the month or year at the start of the period rather than its middle gives 32%; the codes as they stood before the clarifications, 25%; five years instead of eight, 21%; ten years, 33%.
Five outcomes the records couldn't settle. For five projects, a lead (a bankruptcy, a court ruling, an in-service date, a federal refusal, a construction launch) could not be confirmed in a primary record. They keep the outcome they were first coded with, flagged as unverified.
Three model choices the rules left open. The rule said to draw the odds from the count's range, but not on what basis. We kept the line as written, and report the alternatives. Building the range on the 49 projects rather than the 88 observations gives 96.3%, against 96.2%. Centring the draws on the weighted 29%, rather than the raw count (22 of 88 observations, about 25%), gives 96.3%.
The third moves the forecast most. Letting fast-track projects follow each run's approved draw, instead of a fixed 29%, gives 96.4%; the main line keeps them a little above approved projects' average draw, which the rule meant to avoid.
The limits of the sample. It covers projects that made a government list; one that never did is missing. Only British Columbia and Alberta are covered before 2014. Every observation predates the 2025 fast-track laws, which may change the odds either way. The rates are averages across sectors, and most sectors have too few projects to stand alone. They are base rates, not verdicts on any one project.
Sources
- Natural Resources Canada, “Major Projects Inventory 2025: active projects (MPI_2025_Active_Projects_en.xlsx)”, 2025. ↩
- Natural Resources Canada, “Major Projects Inventory geodatabase, English (major_projects_inventory_en.gdb.zip), July 2017 copy”, 2017-07-21. ↩
- Legislative Library of British Columbia, “Major Projects Inventory, British Columbia: collected editions (catalogue item 160920)”, 2026-04-22. ↩
- Government of Alberta, “Alberta Major Projects: historical snapshots export (all dates)”, 2026-10-19. ↩
- Impact Assessment Agency of Canada, “Canadian Impact Assessment Registry - Assessment Inventory (map service layers Completed, In progress, Terminated)”, 2026-10-19. ↩
Changes since publication
A correction fixes something that was wrong. An update records the world moving: odds revised, a prediction settled, data refreshed. Newest first.
The charts, odds tables and other records embedded in this article always show the current data, so some figures may differ from the day it was published. This list records what changed since then.
- Odds
- was
23%now26%
Why: O3 reads the main forecast's probability live, and O2 moved: O2 v1.5 (amendment 4 of our volume model): a share of the oil carried by new pipelines to the coast now counts as reaching the US, most likely 35% (range 10% to 50%); the model counted none before. Evidence: Canada Energy Regulator throughput data by key point show deliveries to Washington State by Trans Mountain's Puget Sound line at about 225,000 to 248,000 barrels a day in January-June 2026, as in January 2024 before the expansion, while loadings at the Westridge tanker terminal rose from about 55,000 to 340,000-529,000 barrels a day; Trans Mountain estimates that 65% of the tankers loaded there since the expansion headed to Asia (Q2 2026 results), and the regulator names the US West Coast as the other destination. The nickel range's high end is read to August 2026 (US share 50.9%, Statistics Canada trade data). O3 follows its official run 78 on that O2 (run 77): 23% to 26%, range 7%-39% to 8%-45%; the count of comparable US deals is unchanged.
In the corrections log: O3: Locked-in preferential terms for the US, 12 November 2026
- Text of the article
Why: O2 v1.5 (amendment 4 of our volume model): a share of the oil carried by new pipelines to the coast now counts as reaching the US, most likely 35% (range 10% to 50%); the model counted none before. Evidence: Canada Energy Regulator throughput data by key point show deliveries to Washington State by Trans Mountain's Puget Sound line at about 225,000 to 248,000 barrels a day in January-June 2026, as in January 2024 before the expansion, while loadings at the Westridge tanker terminal rose from about 55,000 to 340,000-529,000 barrels a day; Trans Mountain estimates that 65% of the tankers loaded there since the expansion headed to Asia (Q2 2026 results), and the regulator names the US West Coast as the other destination. The nickel range's high end is read to August 2026 (US share 50.9%, Statistics Canada trade data). Reworded to what the model now shows: with part of the coast oil counted as reaching the US, the lower build odds from our count of past projects raise the forecast by less than half a point (the section said they raised it, and that each coast project built sends resources overseas rather than south); O2 v1.5 (amendment 4 of our volume model): a share of the oil carried by new pipelines to the coast now counts as reaching the US, most likely 35% (range 10% to 50%); the model counted none before. Evidence: Canada Energy Regulator throughput data by key point show deliveries to Washington State by Trans Mountain's Puget Sound line at about 225,000 to 248,000 barrels a day in January-June 2026, as in January 2024 before the expansion, while loadings at the Westridge tanker terminal rose from about 55,000 to 340,000-529,000 barrels a day; Trans Mountain estimates that 65% of the tankers loaded there since the expansion headed to Asia (Q2 2026 results), and the regulator names the US West Coast as the other destination. The nickel range's high end is read to August 2026 (US share 50.9%, Statistics Canada trade data). The figures follow the new run: P(O2 or higher) 85% to 96%, the two lower rungs 15% to 4%, the median gain US$9.1 billion to US$11.4 billion a year
- Odds
- was
85%now96%
Why: O2 v1.5 (amendment 4 of our volume model): a share of the oil carried by new pipelines to the coast now counts as reaching the US, most likely 35% (range 10% to 50%); the model counted none before. Evidence: Canada Energy Regulator throughput data by key point show deliveries to Washington State by Trans Mountain's Puget Sound line at about 225,000 to 248,000 barrels a day in January-June 2026, as in January 2024 before the expansion, while loadings at the Westridge tanker terminal rose from about 55,000 to 340,000-529,000 barrels a day; Trans Mountain estimates that 65% of the tankers loaded there since the expansion headed to Asia (Q2 2026 results), and the regulator names the US West Coast as the other destination. The nickel range's high end is read to August 2026 (US share 50.9%, Statistics Canada trade data). The O2 odds follow official run 77 of the new version: 85% to 96%; the two lower rungs together 15% to 4%. The reason now says that part of the coast oil reaches the US and that Pacific Link's effect on the odds is small, and names the coast share among the weak points.