What do Tariffs Actually Do?
Concrete Evidence (January 5, 2026)
Happy Monday, and happy New Year. It’s back to work today, even for those lucky enough to get Friday off after the holiday last week.
As we enter the second year of Trump’s second administration, the effects and even legality of tariffs remain big, unresolved issues. The Supreme Court will soon decide if the president’s tariffs are legal under the International Emergency Economic Powers Act. Meanwhile, as new economic data come in and the public gets tired of higher prices, the administration may have to continue adjusting its course.
What do we know about what tariffs actually do? A sweeping new working paper, from two Northwestern-affiliated economists, mines nearly two centuries of American history—with data running from 1840 to 2024, assembled from countless sources—to work out the impact that tariffs have on numerous outcomes.
The big findings: “Tariff increases are consistently contractionary: imports fall sharply, exports decline with a lag, and output and manufacturing activity drop persistently.”
The study identifies 36 different “major tariff events.” The authors studied the historical record of these changes in detail—looking for evidence they were motivated by economic conditions. Such cases raise concerns about “reverse causation,” where the economy is affecting tariffs rather than the other way around.
With that in mind, they focus on 21 of these tariff shifts that seemed motivated by things other than immediate economic concerns, such as international relations or politicians’ ideological commitments to protectionism. These include, for example, the McKinley tariff hike of 1890; the 1913 income-tax law (which also significantly cut tariffs); the 1930 Smoot-Hawley tariffs; the 1947 round of the General Agreement on Tariffs and Trade (designed to promote free trade); and the Trump I tariffs of the late 2010s.
Of course, this “narrative approach” to weeding out causality problems is subjective, and perhaps there’s no such thing as a tariff change totally untainted by what’s happening in the broader economy. But by removing, for instance, “efforts to raise revenue during fiscal stress or to stabilize activity during downturns,” the authors get closer to the ideal experiment, where tariffs are just assigned at random.
Anyway, their modeling of these events suggests bad effects for the economy. For every 1-percentage-point increase in the tariff rate, GDP falls about 0.9 percent. Imports fall 4 percent, though this is an intended effect of the policy. Exports may rise slightly at first but eventually fall too, ending up down around 2 percent. Manufacturing declines as well. The authors note that their findings “complement historical evidence showing that even when protection succeeded in fostering specific industries, it often came at substantial economy-wide costs.”
Interestingly, though, the effect of tariffs on prices is muted. How can that be? The answer is that tariffs increase import costs directly, but also slow the economy, eventually pushing prices back down.
“On the one hand,” the researchers note, “higher tariffs raise the cost of imported goods and intermediate inputs, exerting upward pressure.” And yet “the persistent decline in output, trade, and manufacturing activity generates demand-side disinflationary pressures that dampen the initial cost-push effect.”
Another interesting tidbit is that the impact of tariffs has changed since World War II. In the modern era, prices actually tend to fall, while the decline in exports has become more pronounced.
One factor explaining this change is the end of the gold standard, which allows exchange rates to adjust. When demand for imports falls and Americans spend fewer dollars on foreign goods, the dollar appreciates. The authors write that this, combined with “the associated loss of external competitiveness[,] weigh directly on exports and aggregate demand, placing downward pressure on prices.”
Additionally, the tariff system has become much more reciprocal since America’s early days, with reductions coordinated through free-trade agreements and increases met with retaliation. Tariff changes thus have more indirect consequences than they used to, and more quickly affect exports in addition to imports.
Of course, economists have warned that tariffs are inefficient tax policy more or less forever, while politicians and the public have continued to find them attractive. Whatever the evidence may say, the key decisions remain in the hands of the courts and the president.
Other Work of Note
The Institute of Labor Economics had a ton of interesting working papers lately, including:
Do catalytic-converter thieves respond to metal prices?
How effective is AI tutoring?
Weighting the H-1B visa system toward applicants with higher wage offers would increase ethnic diversity.
At a public flagship university, “an increased share of international student peers reduces the likelihood of majoring in economics for domestic White and Asian men while increasing the likelihood of majoring in economics for domestic men from underrepresented racial and ethnic groups.”
How the CHIPS Act affects employment.
Yet another minimum wage study, with a focus on intergroup gaps.
Elsewhere, an attempt to quantify the growth of Obamacare tax credits.
How K-12 schools can improve threat assessment and prevent more violence.
The Institute for Progress has ideas to accelerate science.
The economics of television content.
Climate change “has already made the United States poorer,” a new study alleges. (But in what counterfactual world did we manage to get as rich as we are without causing climate change?)
“Despite documented increases in years of schooling, the world’s poorest regions still see stagnating outcomes in learning and education quality, potentially creating poverty traps where investments in neither physical nor human capital materialize.”
Cops’ body-camera footage can contain key evidence in some cases, but it’s not like we can review every single interaction to see if cops are behaving as professionally as we’d like. But AI can! This seemed helpful in a couple agencies that ran experiments with it.
See you all next week.