The Promised $200-billion+ Benefits from "Free Trade in Canada" Is Based on Barriers that Do Not Exist

The promised gains aren't based on real inputs: They were generated by numbers from inside the economic formulas themselves - not the Canadian economy.

(Shh … It’s only the models…)

I have been arguing for years that Canada’s projections of massive gains from liberalizing free trade are a fantasy. Part of the reason for this belief was that, as a legislator dealing with businesses and associations over five years about what was required for economic growith, the loudest near-unanimous request was not reducing trade barriers, it was better access to capital. This isn’t figurative: people were literally shouting at moderators that what was required was better access to capital. When there was a call for removal of trade barriers, the argument was always based on the same basic report, which claimed that

• Canada could achieve $200-billion in benefits -

• Based on a single model,

• Run by a single economist,

• In which the trade barriers being “removed” could not actually be measured,

• Which were inferred from a fictional model of a Canada with no internal trade at all.

While deconstructing how the figure of $200-billion came to be generated is important, the fundamental problem doesn’t require any understanding of math or economics at all.

It’s an error in reasoning, because no real-world measurements are going into the models at all. It’s been called “garbage in, garbage out,” or if you’re feeling erudite, Kant and Hume pointed out that you can’t derive synthetic, or empirical conclusions about the world based on purely analytic, mathematical premises.

The even more absurd reason for this is that, by every traditional definition of Canada, we already have free trade. People will say “we need something like the EU” when Canada is more integrated than the EU.

Thes claims of massive benefits have been reported everywhere, cited approvingly, folded into the federal government’s “One Canadian Economy” messaging. King Charles read the round number into the Speech from the Throne.

The latest data - another paper by the IMF - provides an opportunity to analyze the detailed errors.

The latest analysis by the IMF, from January 27, 2026

On January 27, 2026, the IMF published a new analysis — by its economists Federico Díez and Yuanchen Yang, again with the same University of Calgary economist, Trevor Tombe — putting the gain from eliminating internal trade barriers at roughly 7% of GDP, or $210 billion.

These claims have been taken at face value, when just comparing that paper to earlier ones shows the problem.

The 2019 IMF working paper (Alvarez, Krznar and Tombe, Internal Trade in Canada: Case for Liberalization) reported its central figure plainly:

• After stripping out distance and geography, the average “non-geographic” internal trade barrier worked out to a 21% tariff equivalent

• Removing it from goods alone would raise GDP per person by about 3.8%.

The January 2026 analysis reports the same kind of figures:

• The average internal barrier, expressed as a tariff — at about 9%. And;

• It reports the gain from removing barriers at about 7% of GDP.

When you compare these figures, the calculated tariffs have dropped by more than half - from a 21% tariff equivalent to 9%. But the promised payoff nearly doubles - from 3.8% to 7%.

And in the years between the two papers, no province repealed a barrier of this kind, no tariff was discovered, no statute was struck down. The Canadian Free Trade Agreement was already in force in 2019. Nothing in the real world moved in the direction these two numbers moved.

When a bathroom scale reads 200 pounds one week and 90 the next, you do not jump to the conclusion that the person lost weight: you figure there’s something wrong with the scale.

Achieving twice the benefit from removing half the barriers is not a realistic outcome: it tells you that the results are being generated by the math in the model. That is the situation here, which is clear because we now have two readings from consecutive reports.

The mathematical leftovers that explain why the numbers can move in opposite directions:

The reason this is happening is the one I gave last year: the barriers aren’t being measured, because they can’t be. The statistical model is crafted only to measure certain aspects: once the model has explained as much of the trade pattern as it can with distance and borders, whatever is leftover - and unexplained - is labelled a “barrier” and given a number. It is not actual data. It is a residual.

A residual is not a measurement: it is a recognition that something is missing. The size of a residual depends entirely on how the model is built: which sectors are included, which trade elasticity is plugged in, which counterfactual is used.

Add more to the recipe, and you get bigger leftovers: the 2019 version was goods-only and more modest. The 2026 version folds in services, which results in larger numbers, which are then reported in the press as being certain.

Neither is wrong by the other’s standard, because there is no external standard — no actual barrier to check either of them against. There is only the model.

This is exactly what you would expect from a quantity that has no referent in the world. A real measurement of a real thing does not casually halve while the thing it predicts doubles. A number generated by assumptions does precisely that, the moment you change the assumptions.

The papers themselves make it clear that the imaginary barriers are artifacts of the method

You do not have to take my word that the barrier is an artifact of the method. The 2019 paper says so, in its own fine print, three times over, because of the bizarre results being produced: international trade routes are calculated as barriers below zero; missing data is priced as a barrier; and the biggest barriers are found to be on parts of the economy that are not actually traded: local health and education services.

Here they are, explained:

1. The paper calculates that international trade routes are “negative barriers.” When the same method is applied to Canada’s international trade routes, it returns barriers below zero (footnote 24). A barrier of less than nothing is not a barrier; it is the model telling you its leftover bucket has caught something that is not a barrier at all — in this case, the authors admit, a distance measure that overstates geography. If the method invents negative barriers where we can check it, why trust the positive ones where we cannot?

2. Missing data is priced as a barrier. For utilities, the model reports a positive border effect “due to the lack of trade data in this sector” (footnote 22). That means that where the data was absent, the absence itself was scored as an obstacle to trade.

3. Its biggest “barriers” sit on the least tradable things. The highest barriers in the 2019 numbers land on education and health (74%) and utilities (91%) — sectors that are barely traded across provinces for reasons that have nothing to do with policy. You cannot ship a hospital visit from Halifax to Calgary.
   Universities, colleges and schools, and power grids are based on built infrastructure. The common mistake is in treating physical infrastructure and required to deliver the services, at all, are assets that can be broken up and distributed, at no cost.

The 2026 version repeats the pattern, reporting service-sector “tariffs” above 40 and 50%. What the model is measuring there is not a barrier: it’s a building. The simple fact that some things are local by nature. The method cannot tell the difference, so it files non-tradability under “barrier” and hands you a number.

The model generates results that include the impossible - a “negative trade barrier”; where data is missing entirely it is “filled in” and treated as a measurable barrier, and the greatest barrier effects are on parts of the economy are found on activities that are impossible to trade.

In an earlier paper, I compared this to “AI hallucinations” - or confabulations, which is not a metaphor. One of the reasons AI makes things up or gets things wrong is that the databases they draw from reflect statistical relationships between elements. Instead of the gaps - (the leftovers or residuals) being recognized as contrary or missing information, the statistical relationships are used to generate content to fill it in.

The Paper Calculates a Barrier as 1) a Measurement built on 2) a Measurement built on 3) an Inference.

The canonical source concedes on page 8 that “measuring internal trade costs directly is not feasible” which means that the calculation for what is being “removed” was never observed.

Instead, it is inferred in two stages.

First, total bilateral trade costs are backed out of observed trade-flow shares by inverting a gravity relationship (the Head–Ries index), a step that requires importing an assumed elasticity θ before any “cost” exists at all.

Second, that inferred cost is regressed on distance and contiguity, and the "non-geographic, policy-relevant barrier" is defined as the leftover.

The named entity is thus

• The residual

• Of a regression

• Run on an inference.

So it is a residual twice over — and into it loads everything the two stages omit: mismeasured geography, firm-size distribution, demand composition, credit structure.

The construction's own apparatus confirms it’s catching non-barriers. The whole exercise is benchmarked not against any measurable or obervable data, but against a “counterfactual” Canada derived from other models, that attempts to calculate Canada’s economic performance if the provinces were all “autarkies” - zero interprovincial trade, where all residents of a province do business only with one another. This is not a state that exists, or has ever existed in Canada. At no point does that chain touch an observable measurement.

The number that actually was measured

There is one number in the whole exercise that rests on something real, and it is worth holding up against the headline claim of $200-billion.

The 2019 paper looked at the trade agreements that actually happened — TILMA between Alberta and BC, the New West Partnership, the New Brunswick–Quebec and Ontario–Quebec deals — and asked whether trade costs fell after they were signed. They did, a little: the paper associates these real, implemented agreements with cost reductions of 1 to 4% (its Table 4).

What this shows is that every actual liberalization Canada has carried out — signed, ratified, in force — is associated with trade-cost reductions in the low single digits.

The supposed benefits of internal free trade calculated by the IMF assumes that all of the barriers they calculate - whether 9% or 21% - are eliminated in one fell swoop. This is

• A wholesale removal an order of magnitude larger than anything that has ever actually been achieved,

• Applied to a barrier the model cannot locate,

• Costs nothing, and

• Is somehow worth $200 billion.

The last two bullet points explain the appeal: policymakers are under the impression that it will cost nothing and have benefits. It’s too good to be true - and it’s not.

The measured effect of real reform is small. The advertised effect of imaginary total reform is enormous. From the point of view of politics, it’s been said that it’s better to underpromise and overdeliver, but the Throne Speech presented the largest possible figure.

Nothing’s changed, but the number keeps going up

Over the last decade, the calculated benefits of creating a “single market” in Canada has grown steadily.

• 2016 (Albrecht & Tombe): 3–7% of GDP, goods.

• 2019 (IMF working paper): 3.8% of GDP per capita, goods only, from complete elimination.

• 2022 (Macdonald-Laurier Institute): 4.4–7.9%, “$110–200 billion,” “$2,900–5,100 per capita,” once services are added under an assumed mutual-recognition regime.

• 2025: the round “$200 billion” enters the Speech from the Throne and becomes the official federal figure.

• 2026 (IMF): about 7%, “$210 billion.”

Not one step adds an observed or measured barrier: instead, every step up the escalator adds more models of sectors and assumptions. The number is not going up because anyone has found more real-world economic activity: the number grows because the model is asked to do more.

While initial papers at least had the proviso that possible benefits varied very significantly, as the years have passed, the assertion of confidence in the very top number - the optimal possible result - is reported as the likely outcome.

The 2016 paper’s own spread ran from $50 billion to $130 billion — an $80-billion band of uncertainty — and the public conversation kept the top of the next paper’s range and dropped the error bars entirely.

The Realistic Expected Gain to a Typical Household: Approximately Zero

The total figure of $210-billion is portrated as being split equally between all Canadians - that it’s costing every Canadian $5,100, and that the provinces that would benefit the most would be “have not” provinces like Manitoba and Prince Edward Island.

There are a number of reasons why none of this is accurate - and the content of the models explains why.

Here’s the technical mathematical reason: in the Eaton–Kortum/Caliendo–Parro model class the whole lineage of reports uses, the gain is

• A closed-form function of the assumed trade elasticity, θ, and

• The shift in home-spending share (the Arkolakis–Costinot–Rodríguez-Clare result).

What this means in plain language is that, as a calculation, reducing trade costs raises welfare as a theorem:

• Assume θ

• Use it to infer a "barrier"

• Assume the barrier has been removed, and

• The model hands back a positive number by construction.

Growing up in a Household with 14-million rooms

It is also important to highlight the astonishing crudeness and oversimplification these models have when it comes to estimating the impacts on Canadian provinces.

Each province is modelled as one household, with hundreds of thousands or millions of occupants, where only one person makes the decision. Ontario is modelled as a single consumer, Nova Scotia as a single consumer, and so on.

Not a typical Ontarian whose real-world income and property is similar to a large group of others: the model contains exactly one decision-maker per province, and that agent owns all the labour, earns all the income, and consumes all the output of the province.

Rather than track fourteen million distinct Ontarians with different incomes, jobs, debts, and tolerances for moving, it posits one representative household whose choices are taken to reproduce what the aggregate of all those people would do:

• The province's total wage bill is that household's income;

• The province's total consumption basket is that household's shopping.

• When the paper reports "Ontario's real GDP rises 2.9 percent," it means that one household's real income rises 2.9 percent. There is no second person in the province for the first to be richer or poorer than.

In effect, treating an entire province as one household means all of the actual structures of the economy are effectively dissolved, and can’t be measured or detected. Once the province is one household, you cannot ask:

• “Which firms can actually expand into the opened market?”, because you've deleted the firm distribution.

• “Who has the balance sheet to relocate or to bear an adjustment cost?” because the wealth distribution has been deleted;

• “How does a shock propagate through the actual network of supplier–customer relationships?”, because you've replaced the network with a single agent's input–output coefficients;

• “Does the gain accrue to incumbents or entrants, owners or workers, the concentrated or the dispersed;” There is only one agent, so the question is unformulable.

These are not refinements the model left out to make it easier to run, that could then be added back in later. All of the factual details about the real people’s lives, and the benefits and risks of real businesses are defined out of existence by the model’s central assumption - even though the data is publicly available and accessible via both federal and provincial statistics agencies.

Similar errors plague the models’ predictions of where the benefits will land: the model’s migration channel pulls workers out of the provinces with below-average gains (BC, Alberta, Ontario) toward wherever measured productivity rises, on an assumed migration elasticity of 1.5 that was never estimated on Canadian data.

Any claim of “$5,100 per person” figure is two statistical errors stacked on top of each other.

Actual capacity to expand into a newly opened market is concentrated in the largest firms: firm sizes are Zipf-distributed (Axtell 2001).

A Zipf/power-law distribution is scale-free and extremely top-heavy: there is no “typical” size around which items cluster, the way heights cluster around an average. Instead a tiny number of items are enormous and the rest form a long thin tail of small ones, and the large items hold most of the total mass. This is the opposite of a normal (bell-curve) distribution, where the mean is meaningful and extreme values are vanishingly rare. Under a power law the extremes are not rare and the mean is close to meaningless — the average firm size tells you almost nothing, because the distribution is dominated by the few giants at the head.

That is precisely the property the firm-size point in the trade critique rests on. Axtell (2001) showed U.S. firm sizes follow Zipf’s law very tightly — the size distribution is dominated by a small number of very large firms, with a vast tail of small ones. So when a market is opened by removing a trade barrier, the capacity to actually expand into it is concentrated where the size is concentrated: in the head of the distribution. Dividing a projected aggregate gain equally across the population (”$5,100 per person”) assumes a flat or bell-shaped world in which an average is representative. The actual distribution is Zipf — top-heavy and scale-free — so the average is the wrong summary statistic, and the gains accrue in proportion to existing scale rather than per capita. The per-capita figure isn’t just imprecise; it describes a distribution the economy does not have.

The model’s own structure says the gains accrue in proportion to existing size - which means the larger the firms and the higher the concentration of market share and wealth, the greater the gains. Pretending that the benefits will accrue equally to every Canadian household is another fiction, and a fallacy. There is no actual mechanism in the Canadian economy that will deliver those gains equally. Due to pre-existing concentrations of property ownership and income, the greater the concentration, the greater the benefit.

As a consequence, even if you accept the aggregate projection gains, the expected gain to a typical household in Canada is approximately zero.

What’s more, the model imagines that capital will flow from “hot to cold” - supposedly benefiting have-not provinces like Manitoba and Prince Edward Island the most - when in reality, the benefit flows toward concentration.

Further, the model assumes that capital is indestructible and can be reallocated without cost or friction. Capital treated as putty that relocates but is never destroyed. This is not an oversight: it is the model.

Why this matters now

It would be one thing if this were an academic quibble. It is not.

The $200-billion figure is being used right now to argue that Canada’s productivity problem is a trade-rules problem — that the path to growth runs through harmonizing regulations rather than through investing capital in Canadian businesses. In the Globe and Mail today, Andrew Coyne laments of Canada that the provinces have erect hundreds of barriers to trade, when there are virtually none.

I argued last year, and I still believe, that this gets the actual problem backwards: the binding constraint on Canadian productivity is access to capital for real-economy firms, not the paperwork of selling across a provincial line. Other fundamental, massive realities being ignored by policymakers: that Canada’s productivity and investment fell as a direct consequence of the 2014 oil price crash engineered by Saudi Arabia; a pandemic. It is also the result of Canada being in a “balance sheet recession” where Canada’s real households are mired in trillions of dollars in mortgage debt, which is a drag on the economy, and which could easily tip into crisis.

The other omission in all of this is clear from the belief that Canada can get $210-billion in gains at no cost. What’s really missing is what defines capitalism: capital.

But you do not even have to accept my diagnosis to see the problem with the prescription: the goalposts keep shifting. When the calculated barriers were 21%, the prize being dangled was $200 billion. It’s now up to $210 billion, though the barrier has been recalculated at 9%.

A policy target whose size does not move when its own foundation is cut in half is not a measurement of an opportunity. It is a number in search of a justification — and it found one in a model that, by its authors’ own admission on page 8, cannot measure the thing it is named after: “Measuring internal trade costs directly is not feasible.”

The few benefits of regulatory harmonization include truck-safety and securities harmonization, where the absence of common rules has serious consequences.

However, the belief that health care or education can be improved by “efficiencies” is based on the assumption that the system’s grinding crises are due to too much money, and not enough freedom, when the problem - like Canada’s productive industries - is shortages and lack of investment. Canada’s health care system has serious shortages, inadequate staffing and funding, none of which are solved by having the same nurses, doctors and health professionals work in two provinces instead of one.

Change always comes at a real cost in money and time, while the ritual expectation on the part of policymakers is that the change is costless, and that people in systems can drive fundamental reforms from the side of their desk. Poorly funded systems do not save money when patients and hospitals lapse into crises: Crises cost orders of magnitude more than systems that run well, which requires adequate staffing levels, adequate supplies.

Ultimately, the reason why people believe that removing imaginary barriers will deliver benefits is that the models themselves are an expression of a worldview where the nuance of policy has also been flattened. It used to be acknowledged that free trade would have trade-offs - winners and losers, and that the justification for accepting a given deal is that the gains will outstrip the losses, but in aggregate, the whole country will be better off.

This has been flattened into a belief that free trade is a guaranteed way to secure economic growth by removing phantom trade barriers that are the product of unexamined assumptions that are buried in economic formulas.

There are a few fixes specific fixes worth doing, but not because a model promised $200 billion that it cannot find, cannot locate, and now cannot even keep the same size from one paper to the next.

It has to be said, that one of the real sources of Canada’s crises, is that these claims have been taken at face value, and have gone unquestioned to the point that they are the consensus - and little wonder: they are being published by the International Monetary Fund.

At this point the situation is absurd.

Every government in Canada has signed agreements dedicated to removing barriers that do not exist, which will have a likely benefit to the typical household of zero. When the promised growth fails to appear, instead of checking the assumptions, critics double down, and demand that even more imaginary barriers be removed.

The universal consensus compounds the problem: every party supports these measures - NDP, Conservative, Liberal, the Prime Minister, Federal Ministers, Premiers - and more than one Governor of the Bank of Canada has urged this as a priority. Failure to act has been compared to “leaving $20 bills on the street”.

For many such individuals, making a $200-billion mistake is rather embarrassing - which is why a frank admission of fault, and a promise to rectify the situation, is unlikely to bring an end to the hunt to find Canada’s ever-elusive imaginary trade barriers. The more likely outcome, as JK Galbraith put it is that “Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof.”

Though the situation could be summed up by the old saying “We tried nothing, and we’re all out of ideas,” there are real and costly consequences to Canada and its economy from dedicating the resources into trying to stamp out imaginary trade barriers, while real action and investment is delayed.

If you’re feeling up to it and you’re reading this in Canada, please consider sending it to the Prime Minister, your Premier, and your members of the parliament and the legislature and ask them to do their due diligence.

-30-