FEDS Note: A Common Pattern behind Diverging Fortunes: U.S. Tariffs and Exports from Canada and Mexico

FEDS Notes

November 09, 2026

A Common Pattern behind Diverging Fortunes: U.S. Tariffs and Exports from Canada and Mexico

Chris Avalos, Musa Orak, and Avani Pradhan1

The 2025 U.S. tariff increases were followed by sharply different export outcomes in Canada and Mexico. Comparing the 12 months ending in April 2026 with calendar year 2024, Canadian goods exports to the United States fell about 10 percent, while Mexican exports rose about 16 percent, even though Mexico faced a higher average tariff exposure (figure 1). Do these contrasting outcomes mean that tariffs mattered for Canada but had little effect on Mexico, even though both economies depend heavily on the U.S. market?2 To answer this question, we look beneath the aggregates and ask whether more-exposed Canadian and Mexican sectors lost ground relative to less-exposed sectors after the 2025 tariff increases, and whether that weakness was greater in the U.S. market than elsewhere.

In this note, we find a common pattern beneath the divergent totals. Within each country, sectors facing greater tariff exposure experienced weaker exports to the United States after March 2025, when the United States first imposed a 25 percent blanket tariff on imports from Canada and Mexico, which we take as the onset of the tariff shock. Exemptions for goods that comply with the United States–Mexico–Canada Agreement (USMCA) and product-specific tariffs on metals and vehicles left sectors with very different exposure to the new tariffs. A 10 percentage point difference in exposure between more-exposed and less-exposed sectors is associated with roughly 5 percent weaker exports in Canada and 12 percent weaker exports in Mexico. The larger estimated response for Mexico suggests that a given tariff increase weighed more heavily on Mexico's exposed sectors, which also face higher tariffs on average and its exposure is higher and spread across more sectors. However, the difference between the two estimates is not statistically significant. Canada's response is concentrated in steel, aluminum, and copper, the sectors facing the metals tariffs; Mexico's is broader. The same pattern appears when we compare each sector's exports to the United States with its exports to other markets, where tariffs were largely unchanged: for Mexico, a 10 percentage point difference in exposure is associated with about 11 percent weaker exports to the United States relative to exports elsewhere, while Canada's estimate is negative but imprecise. Finally, despite this sectoral weakness, Mexico's exports were resilient in aggregate, reflecting a concentrated surge in machinery exports to the United States, a sector with below-average tariff exposure, rather than strength across its export basket.

Tariff exposure at the sector level

The country-level effective tariff rate already accounts for USMCA exemptions and compliance, but as a single average it conceals how unevenly the new tariffs fall across North American exporters. Because USMCA-compliant goods are exempt, much of bilateral trade carries little or no tariff, while the product-specific tariffs on metals, which carry no USMCA exemption, and on vehicles leave a few sectors far more exposed than the average suggests. Because tariffs passed through almost fully to U.S. import prices and reduced imports of the targeted goods in the 2018 episode (Amiti, Redding, and Weinstein 2019), this variation across sectors is where any effect on exports should most clearly show up. We therefore compare sectors within each country rather than the country-level rates.

Our exposure measure is the sectoral effective tariff rate (ETR) constructed by Eck et al. (2026) for each two-digit HS sector.3 The ETR applies announced tariffs and exemptions to fixed 2024 U.S. import-value weights, so it does not move mechanically with later shifts in trade composition.4 Weighted by 2024 U.S. import shares, it averages 5.4 percent for Canada and 8.0 percent for Mexico (figure 1). Exposure is concentrated. Iron and steel, steel products, copper, and aluminum (HS 72 to 74 and 76) face the sectoral metals tariffs, which carry no USMCA exemption, and have ETRs of 25 to 48 percent; vehicles (HS 87) have ETRs of 10.5 percent for Canada and 17.6 for Mexico; and more than half of the sectors in each country have ETRs below 1 percent.

We estimate the following pooled difference-in-differences regression:

$$$$ \ln X_{cit} = \alpha_{ci} + \tau_{ct} + \beta\,\left(\mathrm{ETR}_i \times \mathrm{Post}_t\right) + \varepsilon_{cit}, $$$$

where $$X_{cit}$$ denotes exports of sector $$i$$ from country $$c$$ to the United States in month $$t$$, $$\mathrm{ETR}_i$$ is sector $$i$$'s effective tariff rate in percentage points, and $$\mathrm{Post}_t$$ equals one from March 2025 onward. The country-by-sector fixed effects $$\alpha_{ci}$$ absorb persistent differences in export levels across sectors, and the country-by-month fixed effects $$\tau_{ct}$$ absorb shocks that hit all of a country's sectors in a given month, such as aggregate U.S. demand, exchange-rate movements, and seasonality. The coefficient of interest, $$\beta$$, measures the percent change in a sector's exports after March 2025 associated with each additional percentage point of tariff exposure, so $$10 \times \beta$$ is approximately the percent change in exports associated with a 10 percentage point higher tariff rate. Because applied rates were near zero before 2025, a sector's ETR is effectively the size of the tariff increase it faced, and because the specification is linear, the estimate is the same whether those 10 points are added to a low or a high tariff. The country-by-month fixed effects absorb anything that moved all of a country's exports together, so $$\beta$$ is identified from differences across sectors within a country and does not measure the effect of tariffs on a country's total exports.

Higher tariff exposure is associated with weaker exports

Table 1 reports estimates of $$\beta$$ from the pooled regression and from the separate country regressions. In the baseline (row 1), exports to the United States weakened more in sectors with greater tariff exposure after March 2025: a sector with 10 percentage points more exposure saw about 7.4 percent weaker exports than a less-exposed sector, relative to their pre-tariff relationship. The country-specific estimates are 4.9 percent for Canada and 11.8 percent for Mexico. Mexico's point estimate is more than twice Canada's, but the difference between the two is not statistically significant (p = 0.23).

Table 1. Export response to 10 percentage points more tariff exposure

Note: Entries are approximate percent differences in exports to the United States associated with 10 percentage points more tariff exposure after March 2025 (10 × β). Country columns are separate regressions with sector and month fixed effects; the pooled column imposes a common slope with country-sector and country-month fixed effects. Monthly data begin in January 2022 except where noted otherwise and end in April 2026 for Canada and May 2026 for Mexico, except in row 9, where the UN Comtrade data run through July 2026 for Canada and June 2026 for Mexico; pooled regressions use the months common to both countries; regressions use 2024 export weights, with standard errors clustered by sector (country columns) or country-sector (pooled). Controls can change sample coverage. *** p < 0.01, ** p < 0.05, * p < 0.10.

Source: Haver Analytics; UN Comtrade; Eck et al. (2026); authors' calculations.

The estimates barely move when we add sector-level U.S. industrial production as a control for sector demand (row 2), add the commodity prices relevant to each sector (row 3), start the sample in 2023 or 2024 (rows 4 and 5), or drop the four months before the onset of tariffs, when exporters front-loaded shipments ahead of the expected tariffs (row 6). The commodity-price control matters most for Canada, where energy and metals prices can move export values independently of shipments.

Row 7 shows where the response comes from. Dropping the four metals sectors leaves Mexico's estimate essentially unchanged, at −11.6 percent, while Canada's estimate becomes small, positive, and statistically insignificant. Canada's estimated response is therefore concentrated in metals: outside those chapters, Canadian exposure is low except for vehicles, so there is little variation left to identify an effect. Mexico's exposure extends beyond metals, notably to vehicles, and its response is correspondingly broader. The pooled estimate without the metals chapters is similar in size but statistically insignificant.

Noting that industrial production (row 2) is an imperfect demand measure, since it may itself respond to trade policy, we next include a second demand control, U.S. imports of the same sector from third countries. These cover every sector, and the pooled estimate remains negative and significant with it, at −6.0 percent (row 8). Canada's estimate is essentially unchanged, but Mexico's estimated response shrinks in magnitude from −11.8 to −5.1 percent. This control is not a clean measure of demand, however. If U.S. buyers replaced Mexican suppliers with third-country suppliers in tariff-exposed sectors, third-country imports rose in precisely those sectors, and conditioning on them absorbs part of the tariff response. The attenuation for Mexico is therefore consistent with supplier switching, although the regression alone cannot separate that from an omitted demand shock. Motivated by these observations, we next ask whether the weakness is specific to the U.S. market.

To do so, row 9 compares each sector's exports to the United States with its exports to other destinations, using UN Comtrade data for both, and asks whether relative exports to the United States also deteriorated more in tariff-exposed sectors. This triple-difference specification, details of which are presented in the appendix, adds sector-by-month fixed effects, which absorb shocks common to a sector across destinations, such as a global slowdown in that sector. Canada's estimate is negative, at −6.8 percent per 10 percentage points of exposure, but statistically insignificant, and the pooled estimate is −8.4 percent. For Mexico, the estimated weakness in relative exports, −11.2 percent per 10 percentage points of exposure, is close to the baseline estimate in row 1. Exposed Mexican sectors therefore lost ground in the U.S. market specifically, not across all of their export destinations.5

The weakness emerged only after the tariffs

The estimates in table 1 average the response over the whole post-tariff period. Figure 2 traces it month by month. For each country, we return to the baseline specification and replace the single post-tariff interaction with a separate interaction for each month, keeping the same fixed effects, weights, and clustering. Each point measures the export differential associated with 10 percentage points more exposure, relative to the average of the months before March 2025; a value of −10 means that exports in a sector with 10 percentage points more exposure were roughly 10 percent weaker, relative to less-exposed sectors, than in a typical pre-tariff month. If the tariffs drove the sectoral weakness, the differential should be flat before March 2025 and turn negative afterward.

The figure shows this pattern. Before March 2025, the differential hovers around zero in both countries, and tests do not detect a linear pre-trend over the preceding three years (p = 0.91 for Canada and 0.83 for Mexico). After March 2025, Mexico's differential turns increasingly negative, reaching about −18 percent by spring 2026, while Canada's deterioration appears later and is less precisely estimated. The monthly version of the destination comparison shows the same pattern: exports to the United States relative to other destinations weaken in exposed sectors only after March 2025, with no detectable pre-trend, although those estimates are less precise (appendix figure A1). The timing is consistent with tariff-related weakening. A causal reading still requires that export growth would not have shifted with exposure for other reasons, such as sector-specific shocks that happened to coincide with the tariffs.

Why Mexican aggregate exports remained strong

The sectoral-level regressions suggest that Mexico's more tariff-exposed sectors performed poorly, even as its aggregate exports grew strongly (figure 1). To reconcile the two, figure 3 decomposes the change in each country's exports to the United States between calendar year 2024 and the 12 months from May 2025 to April 2026 into sector contributions, each equal to the sector's 2024 export share multiplied by its growth rate.

In Mexico, machinery (HS 84), a broad category that includes computers and servers, contributes 17.3 percentage points, more than the entire 15.6 percent increase in aggregate exports. Nearly all of this surge is computers and related units (HS 8471): in U.S. import data, these rose from $48 billion in 2024 to $106 billion in the 12 months through April 2026, accounting for 96 percent of the increase in machinery imports from Mexico. The surge is tied to U.S. investment in data centers for artificial intelligence. The servers are assembled in Mexico largely from imported Asian components, so the domestic value added is small relative to the gross export value.6 Vehicles subtract 2.4 percentage points and electrical machinery adds 1.2 percentage points. Excluding machinery, Mexican exports fall about 2 percent, and the median sector's exports fall 5.2 percent.

In Canada, mineral fuels and vehicles together account for 5 percentage points of the 10.5 percent decline, and the remaining sectors as a group account for most of the rest. The median sector's exports fall 6.4 percent, so the weakness is broad.

Mexico's aggregate strength therefore rests on one sector. It coexists with declines in most sectors and with weaker performance in the sectors facing higher tariff exposure. The decomposition establishes this concentration; it does not say whether the machinery boom will persist.

Implications for interpreting North American trade

Canada's export decline and Mexico's increase mask a common pattern: within each country, sectors facing greater tariff exposure saw weaker exports to the United States. For Mexico, exposed sectors weakened in the U.S. market relative to other destinations, and the aggregate increase rests on machinery. Strong headline growth can therefore coexist with and mask pressure on exposed industries. Conversely, Canada's headline decline is broad across sectors, and only part of it can be tied to differences in tariff exposure, which is concentrated in metals. These patterns matter for the United States as well. The exposed sectors include those supplying steel, aluminum, copper, and vehicles and parts to U.S. manufacturers, and Mexico's server exports are the import side of U.S. investment in data centers.

The estimates describe relative changes in sectoral export values, not the aggregate effect of tariffs on trade or GDP, which also depends on price changes, imported inputs, sales to other markets, and broader demand responses. Some of these adjustments are under way but slow: the Bank of Canada (2026) reports that Canadian businesses continue to seek export markets outside the United States and to reduce their reliance on U.S. inputs. Monitoring sectoral breadth and export destinations alongside the totals gives a fuller picture of how North American trade is adjusting.

References

Amiti, Mary, Stephen J. Redding, and David E. Weinstein (2019). "The Impact of the 2018 Tariffs on Prices and Welfare," Journal of Economic Perspectives, vol. 33 (Fall), pp. 187–210.

Banco Base (2026). "Exportaciones de México 2026: Crecimiento, Cómputo y Retos T-MEC," Grupo Financiero Base, August 10, https://blog.bancobase.com/exportaciones-de-mexico-en-2026.

Banco de México (2026). Informe Trimestral, Enero–Marzo 2026. Mexico City: Banco de México, May 27.

Bank of Canada (2026). "How Canadian Businesses Are Adapting to US Tariffs," in Monetary Policy Report, January 2026. Ottawa: Bank of Canada, January 28.

Eck, Sydney, Trang Hoang, Carter Mix, and Madeleine Ray (2026). "Mind the Gap: Announced versus Implied Tariff Rates in Recent Trade Policy Episodes," FEDS Notes. Washington: Board of Governors of the Federal Reserve System, April 8, https://doi.org/10.17016/2380-7172.4023.

Fajgelbaum, Pablo, Pinelopi Goldberg, Patrick Kennedy, Amit Khandelwal, and Daria Taglioni (2024). "The US-China Trade War and Global Reallocations," American Economic Review: Insights, vol. 6 (June), pp. 295–312.

Appendix

Event study by country (figure 2).

Replacing the single post indicator with a full set of monthly interactions traces the timing:

$$$$ \ln X_{it} = \alpha_i + \tau_t + \sum_{k \neq -1} \beta_k\,\left(\mathrm{ETR}_i \times \mathbb{1}[t = t_0 + k]\right) + \varepsilon_{it}, $$$$

where $$t_0$$ is March 2025 and $$k$$ counts months relative to it. February 2025 ($$k=-1$$) is the omitted month in estimation; figure 2 plots each $$\beta_k$$ relative to the average of the pre-onset coefficients, so that the path is centered on zero before the tariffs. Sector and month fixed effects, 2024 export weights, and clustering by sector are as in the baseline; the static estimates in table 1 collapse the monthly terms into a single $$\mathrm{ETR}_i \times \mathrm{Post}_t$$.

Comparing export performance across destinations.

Did sectors facing greater tariff exposure weaken specifically in the U.S. market, or did they weaken across export destinations? To address this question, we compare each sector's exports to the United States with its exports to the rest of the world, using UN Comtrade data for both destinations.

For each country, we estimate a triple-difference specification across two destinations, $$d \in \{\mathrm{US}, \mathrm{ROW}\}$$. We interact sectoral tariff exposure with an indicator for March 2025 onward and a U.S.-destination indicator:

$$$$ \ln X_{idt} = \alpha_{id} + \tau_{it} + \tau_{dt} + \beta\,\left(\mathrm{ETR}_i \times \mathrm{Post}_t \times \mathrm{US}_d\right) + \varepsilon_{idt}. $$$$

The sector-by-destination effects $$\alpha_{id}$$ absorb fixed differences between a sector's U.S. and rest-of-world sales, the sector-by-month effects $$\tau_{it}$$ absorb any shock common to a sector across destinations (such as a global sectoral slowdown), and the destination-by-month effects $$\tau_{dt}$$ absorb U.S.-wide demand movements. With two destinations, the specification is algebraically equivalent to a two-way fixed-effects regression on the log ratio of U.S. to rest-of-world exports,

$$$$ D_{it} \equiv \ln X_{it}^{\mathrm{US}} - \ln X_{it}^{\mathrm{ROW}} = \alpha_i + \tau_t + \beta\,\left(\mathrm{ETR}_i \times \mathrm{Post}_t\right) + \varepsilon_{it}, $$$$

which is the form we estimate; the two specifications give an identical $$\beta$$. A sector-month enters only when exports to both destinations are positive. The country regressions use 2024 export weights and standard errors clustered by sector. The pooled version uses country-by-sector and country-by-month fixed effects, with standard errors clustered by country-sector.

To trace the monthly path, we replace $$\mathrm{Post}_t$$ with monthly interactions. Figure A1 reports the estimates for a 10 percentage point difference in tariff exposure, centered on the average of the pre-onset coefficients. We use this average rather than a single base month because February 2025 lies on a front-running spike. Negative values indicate that more-exposed sectors weakened more in U.S.-bound exports than in exports elsewhere.

Pre-tariff differential trends are not statistically distinguishable from zero in either country (p = 0.44 for Canada and 0.18 for Mexico), although the monthly estimates are imprecise.

For Mexico, the post-tariff estimate is −11.2 percent per 10 percentage points of exposure (table 1, row 9), close to the baseline estimate. This result supports the interpretation that weakness in Mexico's more-exposed sectors is specific to the U.S. market. Canada's estimate is also negative, at −6.8 percent, but statistically indistinguishable from zero. The pooled estimate is −8.4 percent. Other U.S.-specific sectoral shocks could contribute, and rest-of-world exports may themselves respond to U.S. policy.

1. The views expressed are the authors' own and do not necessarily reflect those of the Federal Reserve Board or the Federal Reserve System. We thank Shaghil Ahmed and Deepa Datta for very helpful comments and suggestions, and Sydney Eck, Madeleine Ray, and Amanda Tinkham for their assistance with effective tariff rates. Return to text

2. In 2024, the United States took 82 percent of Mexico's and 73 percent of Canada's goods exports, more than double the share for any other economy. These shipments equaled 27 percent of Mexican GDP and 18 percent of Canadian GDP. Return to text

3. Eck et al. (2026) apply each product's USMCA compliance rate as reported by the Census Bureau for July 2025, the first month with reliable product-level compliance data, so that a product with 50 percent compliance and a 10 percent tariff receives an ETR of 5 percent. Because compliance likely rose between the March 2025 tariff announcements and July 2025, the ETR partly reflects firms' responses. The sectors receiving the largest tariff increases were also chosen by policy, so tariff assignment is not random. Return to text

4. As a check, we also estimate the same specification with the realized effective tariff rate (RETR), which is calculated as duties paid divided by the customs value of imports. The response remains negative and significant but larger in magnitude, particularly for Mexico. We do not read this as independent confirmation, since realized duties partly reflect firms' own response to tariffs rather than predetermined exposure. Return to text

5. These results strengthen the evidence of U.S.-specific weakness in Mexico's more-exposed sectors. They do not establish that exports elsewhere increased or that tariffs caused the entire baseline response: other U.S.-specific sectoral shocks could contribute, and rest-of-world exports may themselves respond to U.S. policy. Fajgelbaum et al. (2024) show that the U.S.–China trade war changed export opportunities beyond the directly affected destinations, so other markets need not be an unaffected control group. Return to text

6. Mexican exports of computers and related units (HS 8471) rose 178 percent in January to May 2026 from a year earlier and account for about three-fourths of Mexico's export growth in 2026, while imports of the corresponding inputs, chiefly from Taiwan, Vietnam, and South Korea, rose 118 percent (Banco Base 2026, based on INEGI and Banco de México data). Banco de México (2026) likewise attributes the growth of manufacturing exports in early 2026 largely to shipments of computing equipment and parts, with automotive and other exports contributing negatively. The U.S. import figures in the text are from UN Comtrade. Return to text

Avalos, Chris, Musa Orak, and Avani Pradhan (2026). "A Common Pattern behind Diverging Fortunes: U.S. Tariffs and Exports from Canada and Mexico," FEDS Notes. Washington: Board of Governors of the Federal Reserve System, November 09, 2026, https://doi.org/10.17016/2380-7172.4196.

Disclaimer: FEDS Notes are articles in which Board staff offer their own views and present analysis on a range of topics in economics and finance. These articles are shorter and less technically oriented than FEDS Working Papers and IFDP papers.