Research Tariff Impact with Bilateral Trade Data (2026)

Published April 27, 2026 · 1584 words · For researchers

Thirdwatch's UN Comtrade Trade Data Scraper makes tariff-impact research a structured workflow at $0.0015 per record — pull before/after bilateral trade for the tariffed category, surface volume drops, third-country substitution, and pass-through pricing effects. Built for trade economists, policy researchers, and corporate-strategy teams who need quantified tariff impact analyses without commercial trade-database subscriptions.

Why use UN Comtrade for tariff-impact research

Tariff debates are won or lost on data. According to the Peterson Institute for International Economics' 2024 trade tariff database, most major tariff actions of the past decade had measurable but partial impact on bilateral flows — partial because third-country substitution offsets the direct effect substantially. Quantifying that substitution requires bilateral trade data at HS-code level for multiple country pairs, on a year-over-year basis. UN Comtrade is the official source the academic and policy community uses.

The job-to-be-done is structured. A trade economist studying the impact of US tariffs on Chinese steel needs to compare US-China steel imports before and after the tariff plus US-Vietnam, US-South Korea, and US-Mexico steel imports across the same window. A corporate-strategy team in a tariffed industry needs to track substitution flows that affect their competitive position. A policy researcher quantifying the effective burden of a tariff needs both volume and unit-price data. All of these reduce to multi-country bilateral pulls × multiple years × HS-code filtering. The Trade Data Scraper handles each pull in a single call.

How does this compare to the alternatives?

Three options for tariff-impact research using bilateral trade data:

Paid trade databases re-package UN Comtrade with proprietary derivative metrics; the underlying source is what this actor connects to directly. The Trade Data Scraper actor page gives you the structured raw feed; the substitution and pass-through analysis is downstream pandas.

How to research tariff impact in 4 steps

Step 1: How do I authenticate against Apify?

Sign in at apify.com (free tier, no credit card), open Settings → Integrations, and copy your personal API token. Every example below assumes the token is in APIFY_TOKEN:

export APIFY_TOKEN="apify_api_xxxxxxxxxxxxxxxx"

Step 2: How do I pull before/after bilateral imports for the tariffed pair?

Pick a tariff event with a clean date — for example, the Section 301 steel tariffs the US imposed on Chinese imports — and pull bilateral US-China steel imports across the years bracketing the event.

import os, requests, pandas as pd

ACTOR = "thirdwatch~trade-data-scraper"
TOKEN = os.environ["APIFY_TOKEN"]

REPORTER = "USA"
HS_CODE = "7208"   # iron/steel hot-rolled flat products
TARIFFED_PARTNER = "China"
SUBSTITUTE_PARTNERS = ["Vietnam", "South Korea", "Japan", "Mexico",
                       "India", "Brazil", "Germany", "Indonesia"]
BEFORE_AFTER_YEARS = "2018,2019,2020,2021,2022,2023"

resp = requests.post(
    f"https://api.apify.com/v2/acts/{ACTOR}/run-sync-get-dataset-items",
    params={"token": TOKEN},
    json={
        "reporterCountry": REPORTER,
        "partnerCountry": TARIFFED_PARTNER,
        "flow": "imports",
        "hsCode": HS_CODE,
        "years": BEFORE_AFTER_YEARS,
        "frequency": "annual",
        "maxResults": 1000,
    },
    timeout=600,
)
tariffed = pd.DataFrame(resp.json())
print(f"US-China steel: {len(tariffed)} records across {tariffed.year.nunique()} years")

Step 3: How do I pull substitute partners and detect diversion?

Loop the same query across substitute partners; the diversion analysis is downstream.

rows = list(tariffed.to_dict("records"))
for partner in SUBSTITUTE_PARTNERS:
    resp = requests.post(
        f"https://api.apify.com/v2/acts/{ACTOR}/run-sync-get-dataset-items",
        params={"token": TOKEN},
        json={
            "reporterCountry": REPORTER, "partnerCountry": partner,
            "flow": "imports", "hsCode": HS_CODE,
            "years": BEFORE_AFTER_YEARS, "frequency": "annual",
            "maxResults": 1000,
        },
        timeout=600,
    )
    rows.extend(resp.json())

df = pd.DataFrame(rows)
totals = (df.groupby(["partner_country", "year"])
            ["trade_value_usd"].sum().reset_index())
pivot = totals.pivot(index="partner_country", columns="year", values="trade_value_usd").fillna(0)

PRE = "2018"; POST = "2023"
pivot["pre_b"] = pivot[int(PRE)] / 1e9
pivot["post_b"] = pivot[int(POST)] / 1e9
pivot["delta_b"] = pivot["post_b"] - pivot["pre_b"]
pivot["delta_pct"] = (pivot["post_b"] - pivot["pre_b"]) / pivot["pre_b"].clip(lower=0.001)
print(pivot[["pre_b", "post_b", "delta_b", "delta_pct"]].sort_values("delta_b", ascending=False))

The pre-vs-post billion-dollar deltas tell the substitution story. Tariffed partner shows large negative delta; substitute partners show offsetting positive deltas summed across the cohort.

Step 4: How do I separate volume from price effects?

Compute weighted-average unit price per partner per year, then track its trajectory.

df["unit_price"] = df.trade_value_usd / df.quantity.clip(lower=1)
unit_price = (df.groupby(["partner_country", "year"])
                .apply(lambda g: (g.trade_value_usd * g.unit_price).sum()
                                  / g.trade_value_usd.sum())
                .reset_index(name="weighted_unit_price"))
up = unit_price.pivot(index="partner_country", columns="year", values="weighted_unit_price")
up["price_change_pct"] = (up[int(POST)] / up[int(PRE)]) - 1
print(up.sort_values("price_change_pct", ascending=False).head(15))

A tariffed partner showing both a volume drop (Step 3) AND unit-price rise indicates the cost is being absorbed by the importer; volume drop with stable unit-price means the importer simply switched suppliers. The combination is the canonical pass-through-vs-substitution decomposition.

Sample output

A single record from the dataset for one HS code, one year, one country pair looks like this. The tariff impact analysis stitches many such rows.

{
  "reporter_country": "USA",
  "partner_country": "China",
  "flow": "import",
  "hs_code": "7208",
  "hs_description": "Iron/steel hot-rolled flat products",
  "trade_value_usd": 2400000000,
  "cif_value_usd": 2400000000,
  "net_weight_kg": 4800000000,
  "quantity": 4800000000,
  "quantity_unit": "kg",
  "unit_price_usd": 0.50,
  "year": 2018,
  "period": "2018"
}

A typical tariff-impact summary table looks like:

China dropping $1.8B while Vietnam, Mexico, and South Korea collectively gained $2.1B is the textbook third-country-substitution pattern — net imports actually rose despite the tariff.

Common pitfalls

Three things go wrong in tariff-impact studies built on UN Comtrade. HS-code mismatch — the tariff is written at one HS level (typically 4 or 6 digits), and the data analysis must match that level exactly. Aggregating to 2-digit chapters mixes tariffed and non-tariffed sub-products and dilutes the signal. Single-year baselines — using just one pre-tariff year as the baseline is fragile; one-off events (a bad harvest, a strike, a single large contract) can distort it. Use a 2-3 year pre-tariff average instead. Missing transhipment effects — apparent country-level substitution sometimes reflects re-routing rather than genuine new sourcing. Tariffed-country goods can transit through Vietnam or Mexico and be relabelled at customs; this is partially captured in unit-price data (relabelled goods often show artificially high unit prices reflecting transit costs).

Thirdwatch's actor returns unit_price_usd precomputed on every record so the pass-through analysis is straightforward. The pure-HTTP architecture means a 9-partner × 6-year tariff study completes in 15-25 minutes wall-clock and costs about $60 — meaningfully less than any commercial trade-database subscription.

Related use cases

• Track India-China trade flows with UN Comtrade
• Find emerging import-export categories with trade data
• Build a supply chain risk dashboard from trade flows
• The complete guide to scraping business and trade data
• All Thirdwatch use-case guides

Frequently asked questions

What questions does bilateral trade data answer about tariffs?

Three core questions: (1) Did imports of the tariffed category drop after the tariff took effect? (2) Did imports rise from substitute partner countries (third-country substitution)? (3) Did unit prices change, indicating absorption versus pass-through? Each question is answerable from UN Comtrade data with the right query design.

How long after a tariff change does the impact show up?

UN Comtrade publishes annual data with a 12-18 month lag and monthly data with a 3-6 month lag. For a tariff that took effect in 2024, the cleanest before/after analysis uses 2022-2023 (before) versus 2024-2025 monthly (during/after). Don't expect impact analysis on a tariff implemented this quarter; the data needs at least one full reporting cycle to materialize.

What's third-country substitution and why does it matter?

When country A imposes tariffs on imports from country B, importers often shift to country C (a third country) to avoid the tariff while keeping the same goods. This substitution shows up as B's exports to A dropping while C's exports to A rising. The economic effect of the tariff is much smaller than the direct B-to-A drop suggests — third-country substitution often offsets 40-80% of the headline impact.

How do I detect substitution flows in the data?

Pull A's imports from B (the tariffed pair) AND from a list of substitute partners (C, D, E, ...). Compute year-over-year changes for each pair on the same HS code. A substitution pattern looks like: B's exports to A drop 30%, C's exports to A rise 40%, D's rise 25%. The combined diversion to C+D often equals or exceeds the drop from B.

Can I separate price effects from volume effects?

Yes. The actor returns trade_value_usd, net_weight_kg, and quantity per record. Compute unit_price_usd = trade_value_usd / quantity (or use the actor's pre-computed unit_price_usd field). A tariff that maintains import volume but raises unit prices means the cost is being passed through to consumers; one that drops volume without unit-price change is a demand-destruction tariff.

What HS code level is best for tariff analysis?

The HS code level of the tariff itself, ideally. Tariffs are usually written at 4-digit (heading) or 6-digit (subheading) level — match the data extraction to that level. For 2-digit chapter analyses, you'll mix tariffed and non-tariffed sub-products, which dilutes the signal. The actor accepts any HS code level via the hsCode input.

Run the UN Comtrade Trade Data Scraper on Apify Store — pay-per-record, free to try, no credit card to test.