The Anti-Tariff Paper Everyone’s Citing—and What It Doesn’t Really Show
Well, here we go again.
A new NBER working paper is being promoted as making a decisive case against tariffs. While the controversial New York Fed study claimed that recent tariff costs were being borne by Americans, this one claims to show that tariff increases reduce trade, output, and manufacturing.
The paper, by economists Tamar den Besten and Diego R Känzig, is a serious affair. It constructs a narrative series of major U.S. tariff changes from 1840 to 2024 and estimates how tariff shocks correlate with later economic outcomes. Unlike the New York Fed’s anti-tariff disaster, it’s a long-run economic estimate, not a case study of any single presidency or set of tariffs.
But the paper’s findings are being oversold. In the postwar era, the authors’ “clean” episodes are mostly tariff cuts, while most major tariff hikes are excluded as responses to economic conditions. That means the modern “tariff hike” conclusion is largely an inference from liberalization episodes plus two increases from Trump’s first term. So, they’re basically trying to tell us about the economic effects of tariff hikes without actually looking at tariff hikes.
This is the dirty little secret of economists when it comes to empirical studies of tariffs. The “trade liberalization” campaign of the post-World War II era was so successful that they have almost no evidence about what happens when tariffs are raised. So, they’re left guessing or looking at tariff cuts and turning the picture upside down.
The Problem for Trade Economists: Too Many Cuts, Not Many Hikes
The authors identify 35 major tariff events from 1840 to 2024 and classify 21 as “exogenous”—meaning not undertaken in response to fiscal pressures or a slumping economy. In the postwar period, their exogenous list is mostly tariff decreases. It includes eight GATT liberalization rounds, which are all tariff decreases. The only two increases are Trump’s 2018 and 2019 tariffs. They exclude the Nixon surcharge of 1971—the one that led to the creation of the Section 122 tariff authority that Trump just invoked—and the 1975 Ford oil import fees, considering each an “endogenous” response to external imbalances.
That’s a defensible choice. They want to look at hikes that are undertaken as part of a protectionist program or justified by national security or used to correct the distribution of wealth toward the forgotten American. But there’s a problem: from the end of the Second World War until 2018, we did not have any of those. The arc of history or the machinery of progress or whatever was all headed toward lower tariffs in the United States. (There’s also the weirdness that Trump is now describing his tariffs as the response to an external imbalance. And there’s a good case to be made that Trump’s tariffs have always been “endogenous” in the way the paper means it. But we’ll set that aside for now.)
So, that leaves den Besten and Känzig with an awfully thin sample. In the postwar era, the inference about tariff increases leans heavily on what you might call the symmetry assumption: that tariff cuts have mirror-image effects to tariff hikes. This symmetry assumption is pretty standard in econometrics, but it’s not the same as having lots of modern hikes to study. And it’s an assumption the authors never really test.
A simple analogy makes the point. Imagine estimating how the economy responds to a rate hike using mostly rate cuts, then confidently tweeting: “Rate hikes always do X.” You can do it in a linear model. But it’s not the same as actually having many clean hike episodes to examine.
More importantly, the paper never disaggregates the Trump tariff episodes. The 2018 and 2019 tariffs appear as two separate “exogenous” events in their dataset, but they never show us those episodes in isolation or test them against the model’s predictions. They’re blended into an overall estimate that, in the postwar window, is mostly identified from all those tariff cuts.
The authors’ own robustness work underscores how much the postwar results depend on a small number of events. When they drop individual episodes from the postwar sample, the responses swing around and the uncertainty widens sharply—because there simply aren’t many major postwar tariff events to begin with. Yet we still never see whether the Trump tariffs—the only clean modern hikes in their sample—validate the model or cut against it.
Reality Check: 2018, 2019, and 2025 Don’t Match the Pattern
The paper’s estimated “tariff shock” pattern is straightforward in the way it’s being advertised: imports fall on impact, exports fall with a lag, output and manufacturing activity decline, and GDP is lower at the peak effect. But the modern record—the episodes most relevant to today’s argument—show nothing like that happening.
Start with 2018. These were the tariff hikes on aluminum and steel and some Chinese goods. There was lots of screeching about protectionism and the end of the postwar order, but the economy did not roll over. Real GDP growth remained strong year-over-year through the entire year. Real nonresidential investment was very strong. Industrial production rose year-over-year. Imports did not collapse—they continued rising. Exports cooled late in the year, but they didn’t crater. Look wherever you want, you won’t find the automatic broad contraction the paper says should have happened.
Move to 2019. After the May escalation with China, real GDP remained positive year-over-year through the end of the year. Full year, the economy grew a nominal 4.1 percent and a real 2.3 percent. Nonresidential investment stayed positive year-over-year, though slower than 2018. Industrial production turned modestly negative year-over-year—the clearest weak spot in the data. Imports weakened into late 2019, but exports were flat to modestly positive. So, 2019 does show industrial softness. But it does not show an across-the-board collapse in output and trade volumes. It looks mixed: a small manufacturing slump alongside continued aggregate growth and investment that doesn’t implode.
And now we have 2025. Last year’s tariffs are not in the paper because it only runs through 2024, and the last tariff episode it looks at is 2019’s. Still, if the paper’s conclusion held, we should see the contraction it predicts. Guess what? Following the April “Liberation Day” tariffs, real GDP stayed solid through the end of the year and included two quarters of extraordinary growth. Real final sales to domestic purchasers remained positive. Imports turned negative in the second half of the year, the one clear prediction that fits the standard story and also one that many tariff advocates would cheer. But exports stayed positive year-over-year, which runs in the opposite direction of the simple “tariff shock” narrative. Meanwhile, manufacturing hours rose year-over-year, capacity utilization was stable-to-up, and core goods inflation remained modest.
The fact that these three recent tariff hike episodes do not result in the predicted results raises a question: if Trump-era tariff hikes are the only clean modern hikes the authors have, why not show readers whether those episodes match the model’s pattern? Or is it possible that the evidence we’ve just looked at is correct and the only evidence for the contraction comes from holding the picture of tariff cuts upside down?
A New Tariff Regime?
The most important point is one the authors themselves emphasize: regimes matter. They don’t treat tariff effects as constant across history. They stress that tariff transmission differs across monetary regimes and across trade regimes, and they specifically point to monetary-policy responses and foreign retaliation patterns as channels that can change how tariffs transmit through the economy.
This is their framework, not ours. And once you accept it, you can’t casually treat a single historical average as a universal law.
If the mechanism depends on retaliation and monetary-policy reactions, then tariffs used as leverage for deals are not the same policy shock as tariffs raised in a tit-for-tat retaliation environment. If bargaining dominates and retaliation is muted, the “exports collapse” channel weakens. If substitution is faster and investment responds to reshoring incentives, the “output down” story weakens. A tariff regime aimed at rebuilding capacity and extracting market access is not the same object as a tariff change inside a multilateral liberalization regime.
No doubt there will be another tariffmageddon paper next month and another after that and another after that. Maybe one day, the economics establishment will produce one that really lands a blow against Trump’s tariffs. That day, however, did not arrive in February of 2026.
Read the full article here
