Guides
US and Canada border claims, checked against CBP and Census data
Four real US Canada border claims, checked against Census trade tables, CBP statistics and Canadian releases, with what each number actually counts.
What to take away
- Every claim below is graded on one question: does the number count the thing the sentence says it counts?
- Census reports goods and services in separate series, so a "trade with Canada" figure is incomplete unless it names which one.
- CBP counts encounters, not people. One traveler turned back and returning is two encounters.
- A tariff rate is not a cost. Duty collected is the cost figure, and it is a fraction of the import value it is charged on.
- NEXUS enrollment is a stock, crossings are a flow, and the two are published by different agencies on different calendars.
How these four claims were graded
Each claim below is a real shape of claim that circulates in border coverage. Each is checked against the agency that publishes the underlying number, and each gets a verdict: holds up, needs a qualifier, or fails.
The method is the same for all four. Find the table. Read the footnote. Match the unit in the claim to the unit in the table. Then restate the claim using the table's own label and see whether it still reads as true.
A claim that survives that restatement is usable. A claim that only works with the original wording is a labeling problem, not a data problem, and the fix is to change the sentence, not the number.
Claim 1: the US-Canada goods trade balance
The claim: the United States runs a goods trade deficit with Canada of a specific dollar figure.
The check: the US Census Bureau publishes the official monthly and annual goods figures by country. Read the table labels before the numbers. Census separates general imports from imports for consumption, and customs value from free alongside ship value. Those choices move the total.
The verdict: needs a qualifier. The figure is real and checkable, but it is goods only. Add services and the balance moves, sometimes toward surplus. Most political claims about the border are about goods; most economic claims that sound larger include services.
Two more traps sit inside this one. Monthly Census figures are preliminary and get revised, so a claim made in March about February may be checked against a number that has since changed.
And seasonal adjustment changes the sign of a monthly move: a decline in adjusted data can be an increase in the raw data. State which series you used.
Claim 2: northern border encounters
The claim: a stated number of migrants crossed the northern border in a given period.
The check: CBP publishes enforcement statistics covering southwest and northern border encounters, seizures and arrests. Pull the sector-level table, not the national total, if the claim is about New York, Vermont or Washington state.
The verdict: fails as written. Encounters are events, not people. One person who crosses, is returned and crosses again appears as two or more encounters.
A family unit can be counted once as a family and again as individuals in a different table.
The national total is also dominated by the southwest border, so a combined figure can be technically true and still mislead a reader about the north.
The honest restatement names the unit: "CBP recorded N northern border encounters in the period." Whether that is a lot is a separate question, answered by the full series rather than one month.
Claim 3: tariffs on Canadian steel and aluminum
The claim: a tariff on Canadian steel and aluminum costs American consumers a stated amount per year.
The check: CBP collects duties and publishes trade guidance and data through its trade pages at Trade | U.S. Customs and Border Protection. The duty collected figure is the cost figure. The import value is not.
The verdict: fails. The number in the claim is usually the total value of affected imports, and duty collected is a fraction of it. The two are different quantities, and only one is a cost to consumers.
Work it in this order:
- Identify the legal authority for the tariff and its effective date.
- Identify the product scope by tariff classification, not by common name.
- Find the customs value of affected imports from Census or CBP data.
- Multiply by the actual rate, accounting for exclusions and quotas.
- Compare the result to the claim, and state the year and the data vintage.
A blanket claim that a tariff applies to all Canadian steel and aluminum is also wrong at the edges. Tariffs apply to specific classifications, with exclusions, quotas and country-of-origin rules.
If the duty rate is r and the customs value of affected imports is V, the collected duty is r times V, and the reader substitutes their own figures for both.
Claim 4: NEXUS enrollment and crossings
The claim: a stated number of travelers use NEXUS between the United States and Canada.
The check: NEXUS is the trusted traveler program for pre-approved low-risk travelers, with rules set out at NEXUS | U.S. Customs and Border Protection. Enrollment is a stock. Crossings are a flow. A backlog in applications tells you nothing about how many people used the program last month.
The verdict: needs a qualifier. If the claim cites enrollment, say enrollment. If it cites crossings, say crossings, and name the port and the mode.
Biometrics claims follow the same pattern. CBP uses facial comparison at some ports of entry and for some travel modes, described at Biometrics | U.S. Customs and Border Protection. A claim that the border is now fully biometric is false.
A claim that biometrics are used at a named airport may be true, and the port and mode are the qualifiers. Privacy claims turn on what data is collected, how long it is kept and who can see it, and those answers come from published program notices, not from a photograph of a kiosk.
The four failure modes behind all four claims
Every claim above fails in one of four ways: the unit, the denominator, the date, or the source. A reporter sees a CBP release with a large number and drops it into a story without checking what the number counts.
The unit error is the most common. Encounters reported as people, import value reported as tariff cost, enrollment reported as crossings. The fix is to read the table footnote before the headline number.
The denominator error is quieter. A seizure total is a count of what was found, not of what crossed. A rise in fentanyl seizures can mean more smuggling, better detection, or both, and the data cannot tell you which. Any claim that reads a seizure trend as a smuggling trend is unsupported by the source.
The date error is the easiest to catch and the easiest to miss. Border data moves, and a figure from a prior year can be accurate and useless for a claim about this year.
The source error is the one readers notice least. An agency release, congressional testimony and a social media post are not the same kind of evidence. The differences are set out in evidence source types, and they apply directly here.
Matching US and Canadian releases
Canada reports its own trade and border data, and those releases are the check on US figures. Statistics Canada publishes merchandise trade by country and commodity. The Canada Border Services Agency publishes enforcement and traveller statistics. Global Affairs Canada publishes tariff and trade policy notices.
Matching the two countries is not a matter of finding identical numbers. It is a matter of finding the same concept. US import data and Canadian export data describe the same flow, compiled by different agencies under different valuation, origin and timing rules. A gap of a few percentage points is routine.
The exchange rate is a live hazard. Canadian releases often report in Canadian dollars and US releases in US dollars, and a claim converted at the wrong rate can be off by a wide margin. Use the rate the source used, and say so.
The release calendar matters too. A Canadian release for a month may land before the US release for the same month, so comparing a final Canadian figure to a preliminary US figure is not a fair test.
When the two countries disagree sharply, the story is usually the definition: country of origin, re-exports, or valuation basis.
The same discipline applies here as in any other data comparison, and a structured statistics verification checklist keeps you from skipping the boring steps.
Reading the chart that comes with the claim
A line chart of border encounters that starts in a recent low year will look like an explosion. The same data starting a decade earlier may look flat. Neither chart is wrong, but only one is honest about the claim being made.
Check the source line first. If the source is a social media post rather than an agency release, the chart is an illustration, not evidence. The underlying data still has to be checked, and a general statistics verification guide covers how to do that without getting lost.
Border charts add two wrinkles that ordinary charts do not have: two national statistical systems and two currencies. A map is a third. Sector totals shown as state totals is a geography error, and it is common in northern border coverage because the sectors do not follow state lines.
| Problem | What it looks like | How to check it |
|---|---|---|
| Wrong unit | Encounters reported as people | Read the table footnote |
| Wrong denominator | Seizures as a share of all crossings | Confirm the base population |
| Mixed categories | Goods and services combined | Separate the two series |
| Stale data | Preliminary figures quoted as final | Check the release vintage |
| Cherry-picked window | A single month or a peak year | Pull the full series |
| Chart truncation | Y-axis starting above zero | Look at the axis labels |
| Map distortion | Sector totals shown as state totals | Match geography to claim |
These failure modes are catalogued in more detail in statistics reporting problems. The methods behind the gaps between US and Canadian figures are the same ones covered in statistical measures compared.
When a real border number still misleads
Sometimes every number in a claim checks out and the claim is still wrong. This happens when a true figure is used to support a conclusion it cannot carry.
A real number can be real and irrelevant. A seizure total for one port says nothing about a national trend. A monthly trade figure says nothing about an annual policy effect. A program enrollment count says nothing about how many people used the program.
A real number can be real and mislabeled. Calling encounters "crossings" or import value "tariff cost" is a labeling error, not a data error. Restate the claim with the correct label and see if it still holds.
A real number can be real and unrepresentative. One busy month, one sector or one commodity is not the border. Ask what share of the total the cited figure covers.
A real number can be real and old. Date every number, and note the vintage of the table you checked it against.
For the raw enforcement figures themselves, CBP's own stats page is the place to start: Stats and Summaries | U.S. Customs and Border Protection. The agency's stated scope covers trade, travel and enforcement together, not just one of them, as set out at About CBP | U.S. Customs and Border Protection.
Common questions
Where do I find official US trade data with Canada?
The US Census Bureau publishes monthly and annual goods trade by country and commodity. It is the standard source for US import and export values, and each release notes its revisions and data vintage.
Are CBP encounter numbers the same as the number of people who crossed?
No. Encounters count events, and one person can be encountered more than once. Family units and individuals can also appear in different tables, so the totals do not map cleanly to unique people.
How do I check a tariff claim?
Identify the legal authority, the product scope by tariff classification, and the customs value of affected imports. Apply the actual rate, then compare the duty collected to the number in the claim.
Why do US and Canadian trade numbers not match?
They are compiled by different agencies with different valuation, origin and timing rules. Small gaps are normal, and large gaps usually point to a definitional difference rather than an error.
