Trade, by the numbers

Tariffs were supposed to bring factories back. Most factories buy what they tax.

The case for tariffs is that they protect American manufacturing. The catch is in who actually makes things here: most US factories are buyers of steel, aluminum, and parts, not sellers of them. So a tariff that helps the few companies that produce those inputs raises costs for the many that use them. We ran this experiment in 2018, and manufacturing employment fell. The early data from this round is tracking the same way.

The promise

The pitch was a manufacturing rebirth.

When the administration rolled out its sweeping 2025 tariffs, the promise was explicit and large: that walling off imports would pull factories and factory jobs back to the United States.

"April 2, 2025, will forever be remembered as the day American industry was reborn. Jobs and factories will come roaring back into our country."

President Trump, Rose Garden "Liberation Day" tariff announcement, April 2, 2025. (White House transcript via the American Presidency Project)

It is a real and popular idea, and the goal behind it, a stronger domestic industrial base, is one most people share. The question is whether a tariff is the tool that delivers it. To see why it usually does not, you have to look at what an American factory actually is.

The mechanism

Most factories use steel. Far fewer make it.

A tariff on steel or aluminum helps the companies that produce the metal, by letting them raise prices without losing customers to imports. But steel and aluminum are inputs. Thousands of US factories buy them to build cars, appliances, machinery, tools, cans, and equipment. For those factories, the tariff is not protection. It is a tax on their raw materials, one their foreign competitors do not pay.

And the factories that use metal vastly outnumber the ones that make it.

~1

worker producing steel

~80

workers at companies that buy and use steel

Steel-using industries employ roughly 80 times as many workers as steel-producing industries (for aluminum the ratio is even wider). So even a big percentage gain for the protected sector is swamped by a smaller percentage loss spread across the much larger sector that depends on the input. (Econofact, Cox and Russ) Documented

Then foreign governments retaliate, putting their own tariffs on American-made exports, which hits the most competitive US manufacturers and farmers on top of the input-cost squeeze. Protection for a few, higher costs for many, and a smaller export market for the best. That is the structure.

The last experiment

We tried this in 2018. The Fed measured what happened.

This is not a prediction. The 2018 and 2019 tariffs on steel, aluminum, and Chinese goods were studied closely, including by economists at the Federal Reserve, who separated the effect on manufacturing jobs into its three channels.

+0.4%
from import protection (the help for shielded industries)
-1.7%
from higher input costs (factories paying more for metal and parts)
-1.1%
from retaliation (foreign tariffs on US exports)
-2.7%
net effect on manufacturing employment

Comparing more-exposed to less-exposed industries, the small protective boost was outweighed by larger losses from costlier inputs and retaliation. The net effect on manufacturing jobs was negative. (Federal Reserve, Flaaen and Pierce, later published in the Review of Economics and Statistics) Documented

The steel tariffs are the cleanest case. By one count, steel production added about 1,000 jobs, while higher input costs were associated with roughly 75,000 fewer manufacturing jobs downstream. The cost to steel-using industries worked out to around $650,000 per steel job protected. (Econofact; Peterson Institute) Estimate
Retaliation hit exporters and farms hard. China and others targeted US soybeans and goods like Harley-Davidson motorcycles, and the government paid out about $28 billion in farm bailouts over 2018 and 2019 to cover the damage. (Marketplace; Congressional Research Service) Documented

Independent studies reached the same place: the tariffs were passed through almost entirely to US buyers, costing American consumers and firms tens of billions, with little or negative net manufacturing payoff. (Amiti, Redding and Weinstein; Fajgelbaum and coauthors) Documented

This time

And here is what is happening now.

The 2025 tariffs are bigger than 2018's, and the early data is tracking the same mechanism. Start with the headline the promise was about: factory jobs.

about 68,000 fewer

Manufacturing jobs since the January 2025 inauguration, on the official payroll count (12.67 million down to about 12.61 million through May 2026). Revised data and a congressional estimate put the first-year loss closer to 103,000 to 108,000. (BLS via FRED; Progressive Policy Institute / Joint Economic Committee) Documented

And the breakdown by sector is the mechanism in miniature. The metal makers, the protected industries, added jobs. The metal users, the much larger group, lost the most.

In 2025, primary and fabricated metals gained jobs, while machinery, computers and electronics, and transportation equipment shed the most, the third straight year of net manufacturing job losses. That is exactly the pattern the input-cost theory predicts: help the few who make the metal, squeeze the many who use it. (Cato Institute) Documented
Input costs are the reason manufacturers themselves keep citing. The widely-watched factory survey's prices index hit a four-year high in April 2026, with respondents naming steel, aluminum, and tariffs directly. (ISM, via Trading Economics) Documented
The bills show up in earnings. Caterpillar reported up to about $1.5 billion in tariff costs for 2025; Whirlpool absorbed roughly $300 million even though it builds mostly in the US; Deere saw sales fall sharply amid tariff costs and weak demand. (Manufacturing Dive; Reuters via Yahoo Finance) Documented
Retaliation returned too. China cut US soybean purchases to roughly zero from May 2025, and the administration announced a fresh $12 billion farm bailout in December 2025. (Fortune) Documented
And the actual building of new factories is slowing, not surging. Factory construction spending, which had roughly tripled under the 2022 chips and clean-energy laws, has been falling, running well below its 2024 peak. (FactCheck.org; Census via FRED) Documented
The other side, in full

The strongest case for the tariffs, stated fairly.

The honest pro-tariff argument does not deny the weak job numbers. It says the timeline is wrong and the goal is bigger. Here it is at its strongest.

"Factories take years, so judge it in 2027, not 2026." Fair in principle, and some building is genuinely underway: new steel capacity is going up, and the first new US aluminum smelter in decades is planned in Oklahoma, behind the steel and aluminum tariffs, which rest on firmer legal ground. But the sweeping "Liberation Day" tariffs, the ones actually sold as bringing factories roaring back, were imposed through emergency powers that bypassed Congress, and in February 2026 the Supreme Court struck them down 6 to 3. That is the trouble with "just wait": a factory is a multi-decade bet, and almost no company sinks billions into one to dodge a tariff a court has already voided and the next president could drop on day one. Tariffs this uncertain do not pull factories home. They tell companies to wait them out. (Supreme Court ruling via SCOTUSblog; White House)
"The point is strategic capacity, not job counts." A fair goal, and a real one: domestic capacity in steel, semiconductors, and pharmaceutical inputs carries a national-security value a jobs number misses. But it is not the goal he sold. The promise was not "strategic capacity." It was that "jobs and factories will come roaring back." Measuring this by manufacturing jobs is not moving the goalposts. It is using the ones he set.
"Free trade really did hollow out factory towns." Correct, and well documented. The rigorous "China Shock" research found that trade with China cost the US on the order of two million jobs, nearly a million in manufacturing, with lasting damage to the communities hit. That is a legitimate basis for wanting some protection. (Autor, Dorn and Hanson)
"Companies are pledging over a trillion dollars to build here." They are announcing it. The catch is that the largest pledges (chip plants from TSMC and Texas Instruments, for example) were set in motion by the 2022 CHIPS Act, before these tariffs, and announced investment is not the same as a filled job or a poured foundation, which is why factory construction is actually down. (FactCheck.org)

There is also a genuine dissent worth noting: the labor-aligned Economic Policy Institute argues the steel tariffs did help steelworkers and should continue. It reaches a friendlier verdict because it focuses on the protected steel sector itself, rather than on net employment across all the industries that buy steel.

The goal, and what works

The goal is right. There is a tested way to reach it.

None of this means a strong domestic manufacturing base does not matter. It does, for reasons the pandemic and the map both made obvious. When a global chip shortage hit, US automakers had to idle plants, at a cost of roughly $210 billion in lost vehicle revenue in 2021 alone. The vast majority of the world's most advanced chips are made in Taiwan, a few hundred miles from China, a real national-security exposure. And the "China Shock" research found that when manufacturing left, it cost on the order of two million US jobs and hollowed out the towns that lost them. Wanting to rebuild here is the right instinct. (CNBC / AlixPartners; Autor, Dorn and Hanson) Estimate

The question is the tool. And the country just spent three years running a different one that worked: paying companies to build capacity, instead of taxing the inputs they buy.

about $76B → $231B

US factory construction spending roughly tripled from 2021 to its 2024-2025 peak, one of the largest industrial building booms on record, after the 2022 CHIPS Act and clean-energy law. The increase holds even after adjusting for inflation. (US Treasury, on Census data) Documented

The CHIPS Act ($52.7 billion, including about $39 billion in manufacturing incentives) drew roughly $450 billion in announced private semiconductor investment across 28 states, and it produced real plants: TSMC's Arizona fab began making advanced 4-nanometer chips at volume in early 2025, the first leading-edge chips ever made in the US, with Apple as its first customer. (Semiconductor Industry Association; TSMC) Documented
The 2022 clean-energy law's manufacturing tax credit set off a factory wave of its own: roughly 380 battery, solar, and EV plants announced, about half already operating, as actual clean-manufacturing investment more than tripled. (Clean Investment Monitor, Rhodium / MIT) Documented

The difference is structural. A subsidy adds a factory without raising the price of the steel and parts every other factory needs, and without handing foreign governments a reason to retaliate. It grows the pie. A tariff shrinks one part of it to shield another. That is the choice between building capacity and walling off imports, and the construction numbers show which one actually puts up buildings.

Where this approach has its own problems It is not free or flawless. The CHIPS money is about $52.7 billion of public funds plus tax credits, and reasonable people object to the government picking winners. Much of the headline investment is announced, not yet built, and several marquee fabs have slipped, TSMC's second Arizona plant to 2028 and Intel's Ohio fabs further out, with a shortage of skilled workers a real constraint. And it is not a partisan point: the CHIPS Act passed with bipartisan votes. The case here is simply that, on the recent record, paying to build capacity put up factories and a broad tariff has not.
What it comes down to

A worthy goal, and a tool that cuts the wrong way.

Wanting a strong American manufacturing base is not the dispute. It is a goal worth having, and the people who feel that prior trade policy abandoned factory towns are pointing at something real. The dispute is whether a broad tariff is the way to get there, especially when a more direct tool, paying to build capacity, was already working.

The structure of the tool works against the goal. A tariff on inputs helps the small number of companies that make those inputs and taxes the much larger number that use them, then invites retaliation against the exporters who were doing best. When the United States ran this in 2018, the careful measurement found manufacturing employment fell on net. A year into a larger version, the official jobs count is down, the factories that buy metal are losing more than the ones that make it are gaining, input costs are at multi-year highs, farmers are being bailed out again, and the building of new plants is slowing.

The case for the tariffs now rests almost entirely on the future: that capacity being seeded today will become jobs in 2027 and 2028. That bet may yet pay off, and it deserves to be judged on its own timeline. But the specific claim made in 2025, that the factories were already roaring back, is not what the numbers show. So far they show the opposite, for the reason they showed it last time: most American factories are on the paying end of a tariff, not the protected one.

Sources

Where this comes from.

Load-bearing numbers come from the Federal Reserve, the Census Bureau, BLS, peer-reviewed economics, and nonpartisan researchers. Where a number is a model estimate rather than a head count, it is labeled.

The promise and the mechanism

The 2018-2019 evidence

The 2025-2026 data

The other side

This page assesses the claim that tariffs revive US manufacturing, using the 2018-2019 record and data through May 2026 (the latest available; the June jobs report lands in early July). Job-effect figures from the 2018-2019 studies are model-based estimates comparing more- and less-exposed industries, not direct head counts. The 2026 factory survey also rebounded into expansion on orders even as hiring stayed soft and input prices hit highs, and some cost pressure in 2026 reflects energy and other factors, not tariffs alone. Where estimates differ or evidence is contested, the page says so. Corrections welcome.