Buy Canadian: What Actually Works — We Parsed 1.3 Million US Customs Records

What a heart attack in Nelson, B.C. has to do with a trade war

In August 2026, a man named Ben called us from Nelson, British Columbia. He had had a heart attack. His licence was suspended for three months, his job is twenty minutes away, and an electric bike had become the difference between working and not working.

He had ordered one on Amazon. Two hours later he found a Canadian retailer, and he cancelled the Amazon order on principle — a cancellation that took him two days, a seller who offered him an off-platform side deal, and a phone call to demand his own money back.

The Canadian option cost about $225 more. Ben works twenty hours a week. In his words, on the record: "I only work twenty hours a week, and I'm grateful for that job because it's not easy to find work in a small mountain town."

Here is the part nobody mentions. Ben's paycheque comes from American tourists. Half his store's summer customers cross the border from Washington State, and when we asked whether the trade war had changed that, he said: "Not really."

He boycotts American commerce with his own wallet while his wages are paid out of American wallets — and Americans kept coming: Statistics Canada records the collapse running one way only. That is not hypocrisy. That is what a trade war actually looks like from the ground, and it is why the standard advice fails the people it is aimed at.

What Canadians actually did — measured by the Bank of Canada

In February 2026 the Bank of Canada published Assessing the Buy Canadian movement one year later, by economists Olga Bilyk and Jacob Dolinar. They used border-crossing counts and the NielsenIQ Homescan panel — roughly 10,000 households a month, barcode-level. It is the most rigorous look at what Canadians actually did, and the results are lopsided.

Ten million fewer trips. A 25% collapse — 30% on land crossings. That is one of the largest voluntary consumer shifts in recent Canadian history, and Canadians did it without being asked twice.

Then look at groceries: two percentage points — and that is the share of food spending specifically. The Bank’s authors add a caution that matters. The biggest movements were in coffee and fruit juices, and they note the shift could also reflect Canada’s own counter-tariffs, which raised prices on some US food products and may have prompted shoppers to look for cheaper alternatives. Orange juice is the example they give. They do not claim to have separated price response from patriotism, and neither do we.

Outside groceries, in the Bank's own words: "we see no comparable shift in spending patterns." They attribute it to a shortage of Canadian alternatives.

Who wrote the instruction — and why it said "shop"

Every mass movement gets its instructions from whoever can afford to publish them. In this case, that was retailers.

When "buy Canadian" surged, the grocery chains put maple leaves on their shelves. The CBC's framing at the time was blunt: buying Canadian was a matter of pride for shoppers, and for the major chains, an opportunity.

Then the labels started appearing on things that were not Canadian. Between November 2024 and February 2026 the Canadian Food Inspection Agency recorded 78 country-of-origin violations across all Canadian retail — not by any one chain. In January 2026, two Loblaw-owned stores were fined $10,000 each. One case involved American-made President's Choice broccoli slaw carrying maple-leaf decals and a "Product of Canada" shelf tag.

Then the labels came down. Loblaw stripped country-of-origin information from produce signage; Sobeys dropped the maple leaf symbol.

Then, on 26 August 2026, Loblaw CEO Per Bank announced on LinkedIn that the company would reintroduce country-of-origin labels in its produce aisles and bring back the “T” symbol marking tariff-affected items. The maple leaf symbol, which Loblaw had continued to use throughout, stays. He wrote that “Loblaw will not benefit from tariffs. Where tariffs increase our cost, any resulting increase on our shelves will reflect that impact — penny for penny.” Sobeys said it would add provincial-flag signage; Metro said it already prioritises Canadian products.

Two corrections we owe the record. It was Sobeys, not Loblaw, that dropped the maple leaf symbol in stores — and the CFIA subsequently investigated advertising practices overseen by Sobeys’ head office and closed the file without fines after the company took corrective action. Metro has no recorded violation in any of this and appears here only because it was asked to comment.

We are not calling anyone evil. Grocery retail genuinely runs on thin margins — the Competition Bureau's Retail Grocery Market Study calls it “a low-margin business” and found gross margins rose only one to two points since 2017. (An earlier version of this article attributed a “2–5%, lowest of any consumer-facing industry” figure to that report; it is not in the report and we have removed it.) We took the wider cost picture apart in why Canada is so expensive. That is a real constraint.

But shelf labelling costs almost nothing, and the labels tracked the marketing weather: up when the movement sold, misapplied 78 recorded times, down when enforcement started, back up when the political pressure returned. Our view: that reads as a campaign rather than a commitment. We looked at how the retaliation itself was walked back in why Canada dropped its retaliatory tariffs.

How to tell if a product is actually Canadian

The most common question Canadians ask on this subject is some version of "how do I know if this is Canadian?" The answer is a legal definition, and it is stricter than most people assume.

The Competition Bureau's guidance (last updated 14 June 2026) sets two different bars:

So "Made in Canada" can legally mean 49% imported. That is not a loophole; it is the published rule, and the qualifying statement is there precisely so you can see it. The Bureau's guidance covers non-food goods — food labelling falls to the Canadian Food Inspection Agency.

The services half — what Canadians actually did to it

Canadians sent the United States $92.3 billion in services in 2025. That flow produced a $27.7 billion American surplus, which offsets 57% of the $48.3 billion goods deficit. Services are the half of the trade relationship almost no boycott advice mentions — and the half where Canadians landed their single hardest blow without anyone calling it a boycott.

Two things fall out of that chart, and both cut against the easy story.

First: the biggest single line is not digital at all. Other business services — management consulting, legal work, engineering, including advertising and related services — is 28.8% of the total. No consumer can cancel any of it. The categories a household subscription actually lands in, telecommunications/computer/information services plus intellectual-property charges, come to about $18.2 billion. Real money. Not $92.3 billion.

Second, and this is the finding: Canadians already ran the largest services boycott available to them, and it worked. Travel is a service. It is 19.8% of that total, and in 2025 it fell $3.4 billion — down 15.6%, the steepest decline of any category in the account. Air passenger fares fell another 10.6%. Ten million fewer trips showed up in the American national accounts exactly where you would expect.

Why Ottawa cannot do this for you

Twice, Canada tried to make American technology companies pay, and twice it stopped under trade pressure.

The Digital Services Tax — a 3% levy whose named targets included Amazon, Google and Meta — was rescinded on 29 June 2025, hours before the first payments were due, two days after the President suspended trade negotiations over it. Forgone revenue: roughly $7.2 billion over five years.

Then, in a 17 July 2026 filing to the Federal Court of Appeal, Canada confirmed it will eliminate the base contribution requirement on foreign streaming services and replace it with government funding — meaning the Canadian taxpayer now pays for Canadian content instead of the streamers.

And the government's own Buy Canadian procurement policy is legally boxed in. Noah Fry, a Killam postdoctoral fellow at Dalhousie who studies public procurement, found that under the $25 million threshold only six contracts out of roughly 56,000 annually were not already going to Canadian suppliers — narrowing to two once you isolate genuinely competitive bids. Trade agreements forbid most origin discrimination. We traced the wider dependency in what happens if America stops being Canada's friend.

What we found in the American customs file

Rather than trust anyone's list, we downloaded the United States' own export record — every state, every product, every destination — and parsed 1,309,098 lines. Three things fell out that we have not seen published anywhere.

1. A tenth of the trade cannot be traced to any state

$38.85 billion — 11.6% of everything the US exported to Canada in 2025 — carries the state code "XX," attributable to no state at all. Every state-level analysis of this trade, including the ones cited by legislators, is working with a denominator missing more than a tenth of the total.

2. The data credits farm exports to the wrong states

The series records the state a shipment moved from, not where the goods were grown or made. For bulk crops railed to export terminals, that produces results that look absurd once you check them against production:

The distortion is not universal, and the exceptions prove the mechanism. Apples: Washington is credited with 77% and grows roughly 60–67% — close. Canola: North Dakota is credited with 66% and grows 81–83% — close, and slightly under. Both are goods that ship direct rather than transiting a distant port.

3. Some "American" food is not American

$517.9 million of bananas were exported from the United States to Canada in 2025, credited to California, New Jersey and Delaware. The United States grows essentially no commercial bananas — it imports about 9.2 billion pounds a year, against a Hawaii crop of 4.7 million pounds. Delaware is a fruit port, not a plantation.

The same applies to $428 million of prepared nuts (cashews, almonds, pistachios and Brazil nuts — several of which are not grown in the US at all) and much of $466 million of roasted coffee (the roasting is American value-add; the bean is not).

4. Consumer pressure works on some states and not others

Across all 50 states there are 97,761 distinct product lines flowing to Canada. Classifying them by whether an individual could plausibly decline them at retail produces a spread most people would not expect:

Michigan sells Canada $23.25 billion and only about 6% of it is anything a person could choose not to buy. The rest is automotive components and industrial goods. New Jersey's figure is seven times higher. A consumer boycott aimed at one delegation is close to symbolic; aimed at another it touches four dollars in ten.

The map: which state absorbs it, and who represents them

This is the part no grocery chain will ever print for you. Using the American customs file, we can say for any everyday product which state ships the most of it to Canada — and from the current Congressional roster, exactly who represents that state in the Senate.

Two weightings matter, and they answer different questions. Share of US supply tells you who absorbs it if you decline — of everything America ships Canada in that product, how much leaves from that one state. The second, in our published data files, is what share of that state's total Canada-bound trade the product represents — how much the state actually depends on it.

Food & drink

Household & personal

The script, so you can check us

This is the parser we ran against the US Census file. Download STHS6M2512.ZIP from the Census State Exports HS6 page, unzip it, and run this against STHS6R2512.txt. It should print $333,619,712,376. If it does not, we are wrong and we want to know.

#!/usr/bin/env python3
"""
Parse US Census "State Exports HS6 (Origin of Movement)" fixed-width customs files.

PROVENANCE
  Source : https://www.census.gov/trade/downloads/2025/state_exp/hs6_om_m/STHS6M2512.ZIP
           (listed on https://www.census.gov/foreign-trade/data/STHS6M.html)
  File   : STHS6R2512.txt  — Dec 2025 monthly file; YTD columns = full calendar year 2025
  Pulled : 2026-08-27
  Rows   : 1,309,098

RECORD LAYOUT (inferred from the file, then VALIDATED against an independent published total —
see validate() below). Fixed width:
    [0:6]   HS6 commodity code
    [6:10]  Country code (Canada = 1220)
    [10:12] State of origin-of-movement (USPS 2-char)
    [12:18] YYYYMM
    then 14 right-aligned numeric fields, width 15 each:
      1-7  monthly : ALL_VAL_MO, AIR_VAL_MO, AIR_WGT_MO, VES_VAL_MO, VES_WGT_MO, CNT_VAL_MO, CNT_WGT_MO
      8-14 YTD     : ALL_VAL_YR, AIR_VAL_YR, AIR_WGT_YR, VES_VAL_YR, VES_WGT_YR, CNT_VAL_YR, CNT_WGT_YR
    -> ALL_VAL_YR (field 8) is the calendar-year total value in USD.

Values are nominal USD, origin-of-movement basis (the state the goods were shipped FROM, which is
not necessarily where they were produced — a known limitation of this series, state it in print).
"""
import sys, collections

CANADA = "1220"
VAL_START, W = 18, 15

def rows(path, country=None):
    with open(path, "r", errors="replace") as fh:
        for line in fh:
            if len(line) < VAL_START + W * 14:
                continue
            cty = line[6:10]
            if country and cty != country:
                continue
            vals = [line[VAL_START + i*W : VAL_START + (i+1)*W].strip() for i in range(14)]
            try:
                all_val_yr = int(vals[7] or 0)
            except ValueError:
                continue
            yield {"hs6": line[0:6], "cty": cty, "state": line[10:12],
                   "time": line[12:18], "all_val_yr": all_val_yr}

def validate(path):
    """Sum ALL_VAL_YR for Canada across all states and compare to USTR's published
    US goods exports to Canada 2025 = $333.6 billion. A match confirms the layout."""
    total = 0; n = 0; states = collections.Counter()
    for r in rows(path, CANADA):
        total += r["all_val_yr"]; n += 1; states[r["state"]] += r["all_val_yr"]
    print(f"Canada rows            : {n:,}")
    print(f"Sum ALL_VAL_YR (Canada): ${total:,}  = ${total/1e9:,.1f} B")
    print(f"USTR published 2025    : $333.6 B  -> delta {(total/1e9-333.6)/333.6*100:+.1f}%")
    print(f"Distinct state codes   : {len(states)}")
    return states

if __name__ == "__main__":
    validate(sys.argv[1])

# ---------------------------------------------------------------------------
def dependence_and_concentration(path):
    """For every state: exports to Canada, exports to the WORLD, Canada's share,
    and how concentrated the Canada-bound basket is (top-1 and top-3 HS6 share).
    Concentration is Milad's apple test: a state carried by one product has no shock absorber."""
    to_can   = collections.Counter()
    to_world = collections.Counter()
    prod     = collections.defaultdict(collections.Counter)   # state -> hs6 -> $ to Canada
    for r in rows(path):
        st, v = r["state"], r["all_val_yr"]
        to_world[st] += v
        if r["cty"] == CANADA:
            to_can[st] += v
            prod[st][r["hs6"]] += v
    out = []
    for st, can in to_can.items():
        world = to_world[st]
        top = prod[st].most_common(3)
        t1 = top[0][1]/can*100 if can and top else 0
        t3 = sum(v for _, v in top)/can*100 if can else 0
        out.append({"state": st, "to_canada": can, "to_world": world,
                    "canada_share": can/world*100 if world else 0,
                    "top1_share": t1, "top3_share": t3,
                    "top": top})
    return out

Our own books, since we are asking you to inspect everyone else's

An article arguing that the "buy Canadian" instruction was written by people who sell things is worthless unless the people writing it disclose that they sell things too. So here is our ledger.

Zeus is a retailer, and we charge a premium over the cheapest way to buy the same product. That premium pays for a Canadian business staying in the transaction — the phone that gets answered, the warranty someone actually processes, the hour we spent on the call that produced this article. When Ben asked whether we would close the $225 gap between us and a cheaper option, we said no. He is choosing a dealer closer to home, partly because we told him to. You should weigh that premium against what this article ranks it as: fifth of five, the costliest action with the smallest national effect.

None of our bikes are made in Canada. Neither are almost anyone's. We wrote a separate, honest accounting of that in our guide to Canadian-made electric bikes, including which brands assemble here and which merely say so.

We could not follow our own advice completely — but we did move what we could. Our store runs on Shopify, an Ottawa company. And the single biggest line item we actually switched was video and image production.

Every photograph, cover and video Zeus publishes used to run through American creative tooling. It now runs through Playcut, built by Blox Labs Inc. — a Canadian-headquartered team, Toronto-registered with roots in London, Ontario. That is not a token switch. For a small retailer, creative production is one of the few recurring software costs large enough to matter, and it is one of the rare categories where a genuinely Canadian option exists at the same quality. It let us build a consistent visual identity across hundreds of articles without a studio, a crew, or a monthly cheque leaving the country. If you run a Canadian business and you are looking for the switch that is actually available rather than theoretically available, that is ours, and we would make it again.

Here is what we could not switch. Search, social, communications and analytics all still run on American platforms. We looked hard at each one. In those categories there was either no Canadian equivalent at all, or none we could move to without breaking the business — which is exactly the dependency this article is about.

The services map: which app, which state, which senators

The state map above is for groceries, which this article ranks fourth. Here is the same map for the action ranked first — and it is far more usable, because you do not need a label to tell you what you subscribe to. You already know.

Sixteen American services, where each company is headquartered, who represents that state, and what is on the public record about that company and the current administration.

Three services people get wrong

What actually works, ranked

Every row is ordered against the Bank of Canada’s behavioural findings and the USTR trade balances, on four measurable factors: how much of the dollar leaves Canada, whether it recurs, whether the company reports the loss to shareholders, and what it costs you.

Frequently asked questions

Does buying Canadian actually work?

Partly, and unevenly. The Bank of Canada measured a 25% drop in trips to the US and roughly a 2-percentage-point shift in grocery spending, with no comparable movement in other categories. The pattern is that Canadians act when the action is free and an alternative exists.

What does “Made in Canada” legally mean?

At least 51% of total direct production costs incurred in Canada, the last substantial transformation performed in Canada, and a qualifying statement disclosing imported content. “Product of Canada” requires at least 98%. Those are Competition Bureau thresholds for non-food goods; food falls under the Canadian Food Inspection Agency.

What is the single highest-impact thing I can do?

On the evidence: cancel an American subscription you are not using. Nearly the whole dollar leaves Canada, it repeats every month, it is visible in the company’s own reporting, and it refunds you rather than costing you.

Is it worth paying more for a Canadian retailer?

Sometimes — but be clear about what you are buying. If the product is manufactured overseas either way, the premium buys a Canadian business staying in the transaction, not Canadian manufacturing. That can be worth it. It is a different thing from what most people think they are paying for.

Why doesn’t the government just require Buy Canadian?

Trade agreements largely forbid origin discrimination in procurement. Research by Noah Fry at Dalhousie found only six contracts out of roughly 56,000 annually were not already going to Canadian suppliers — two, once competitive bids are isolated.

How can I tell which US state a product came from?

For most goods you cannot — the label reads “Product of USA” and nothing more. We tried to solve this with meat labels and it does not work: a full-text search of the Safe Food for Canadians Regulations returns no requirement for a foreign establishment number on a Canadian label. Canada requires a country name and the foreign inspection mark, and only the last processor’s mark may appear on immediate packaging.

What to do on Monday

Open your phone's subscription list. Most people find one or two they forgot they were paying for. That single action outranks a month of careful label-reading at the grocery store, costs you nothing, and gives you money back.

Then, if you want the harder version: the ranked action ledger and the state-by-state product tables are already published as a free printable field guide — open it here. No email, no signup, free to copy and share.

What this article cannot tell you

Every dataset here has a boundary, and a piece that hides them is not evidence, it is advocacy. These are ours.

Corrections

Every correction to this article is listed here with its date, rather than made silently.

Sources

Every source below was retrieved and read on 27 August 2026.

• US Census Bureau. State Exports HS6 (Origin of Movement), December 2025 file, calendar-year totals. 1,309,098 records, parsed in full. Canada-bound total: $333,619,712,376.
• US Census Bureau. Schedule B export commodity classification, 2026 edition.
• Office of the United States Trade Representative. Canada country page, 2025 goods and services balances.
• US Bureau of Economic Analysis. International Services, Table 2.3, released 7 July 2026 — services to Canada by type.
• Bank of Canada. Olga Bilyk and Jacob Dolinar, Assessing the Buy Canadian movement one year later, Sparks at Bank article 2026-6, 26 February 2026. Views are the authors’, not the Governing Council’s.
• Statistics Canada. Canadian-resident travel to the United States: A year in review.
• Competition Bureau of Canada. Made in Canada claims, modified 14 June 2026.
• Competition Bureau of Canada. Retail Grocery Market Study, June 2023.
• Department of Finance Canada. Rescission of the Digital Services Tax, 29 June 2025.
• Federal Court of Appeal filing, 17 July 2026, on the elimination of the base contribution requirement for foreign streaming services — as reported and documented by Michael Geist.
• Federal Election Commission. Trump-Vance Inaugural Committee, committee C00894162, Form 13 as amended.
• US Securities and Exchange Commission. EDGAR company search — principal executive office addresses read from each registrant’s most recent 10-K or 10-Q cover page, June to August 2026. (SEC rate-limits automated requests; the link opens normally in a browser.)
• Executive Order 14352, Saving TikTok While Protecting National Security, 90 FR 47219.
• USDA National Agricultural Statistics Service. Crop Production 2025 Summary, 12 January 2026.
• Noah Fry, Killam Postdoctoral Fellow, Dalhousie University. “Why the Buy Canadian strategy won’t cost — or deliver — much”, Policy Options, 10 June 2026.
• Canadian Food Inspection Agency country-of-origin enforcement records, as reported by CBC News.
• Ben, Nelson, British Columbia. Recorded telephone conversations, 24 and 26 August 2026, quoted with his consent.