One Year of Tariff Analysis: What We Got Right, What Changed, and What We Learned

One year ago today, on April 2, 2025, the administration announced the most sweeping tariff increases since Smoot-Hawley. At the time, The Budget Lab projected that the resulting suite of tariff policies would increase consumer prices by 2.3%, costing the average American household $3,800 per year.

Looking back after a year of higher tariffs, we aim to re-examine this estimate of the effect of tariffs on prices. The 2.3% estimate we reported last year is large relative to more recent numbers released by academics, policy analysts, and our own, more up-to-date analysis. It is possible this is because not enough time has passed yet for the full burden of tariffs to manifest in the economy; it may take longer than one year for these effects to be felt due to issues like stockpiling of goods and policy uncertainty. However, if one had assumed that the full impact of tariffs would occur within one year and appear entirely as higher prices rather than through some other mechanism (say, via reduced nominal incomes) then our 2.3% estimate implies almost 80% of inflation over the past year is due to tariffs.

This presents us with an opportunity to reflect on what we got right and wrong in the modeling of tariffs over the last year. We can contrast our original estimate with more recent work done by Cavallo, Llamas, and Vazquez (2025) and Dvorkin, Leibovici, and Santacreu (2025) that finds the “true” price effect to be lower, in the range of 0.5% to 1.0%. It is worth noting that even a retrospective attempt to identify this type of causal estimate is fraught. However, these efforts nonetheless constitute our best guess, given the state of the literature, of what a reasonable effect might be.

So, what explains the gap between our estimate from April 2, 2025, and what do we know today about prices? And what does this tell us about the accuracy of our tariff modeling efforts? We decompose the gap into four major components, summarized in Figure 1 below.

First, our estimate from last year assumed that the policies enacted on April 2nd would stay in effect in perpetuity—and be understood as such by market participants—which is a standard modeling assumption made by government scorekeepers that helps to clearly ground policy analysis. This is of course not what occurred, as tariff policy changed often after April 2025, which represented the peak of tariff rates last year. We therefore re-estimate our initial model using our understanding of tariff rates as released in our State of Tariffs (SOT) reports throughout the year. This reduces the price effect from 2.3% to about 1.8%, a 0.6 percentage point drop (when comparing unrounded estimates).1

Second, the precision with which we characterize and measure tariff rates in real time has improved, culminating in our just-released Daily Tariff Rate tracker. Critically, while our estimates of tariff rates have been fairly accurate over the course of the year, our treatment of the timing of those changes, for example, consistency in our handling of tariffs that are merely announced versus those that are implemented, has changed. Adjusting for both the actual observed rates and the timing of those rates, we re-estimate the effect of prices using our best estimate of the average tariff rate over the course of April through December of this last year, which lowers the estimated effect of tariffs on prices to 1.4%, a further 0.3 percentage point reduction.

Third, we have made several methodological improvements over the course of the year, including a more granular approach to estimating price impacts. When we reproduce our estimate using our updated model, we find a price effect of just under 1.4%, indicating that methods do not explain much of the gap between our initial estimate and current estimates. Even though our updates have improved our estimates, particularly when it comes to identifying price effects in detailed consumer spending categories, these adjustments had offsetting effects on the overall estimated price effect.

Even after these adjustments, our estimate remains above the approximately 1% effect estimated by Cavallo, Llamas, and Vazquez (2025). We therefore turn to examine the economic assumptions that are embedded in our model.  We focus on one of the most impactful: the assumed share of the cost of tariffs that are “passed through” to consumers versus those that are borne by foreign exporters, US importers, or workers. (Put differently, we assume the Federal Reserve fully “looks through” the tariffs, allowing the price level to rise rather than letting nominal incomes for firms and workers fall.) We therefore present several iterations of our model where we vary the passthrough rate consistent with our modeling of the most up-to-date pricing data. Depending on the passthrough assumption, we get a short-run price effect of between 0.55 and 1.05 percent.

Other economic assumptions could also be important, though we do not analyze them in detail here. For example, the adjustment process for the consumer price level could take several years to play out, meaning that, at one year into this tariff regime, it’s still too soon to fully evaluate price impacts. We discuss this and other similarly important assumptions in the “Other Open Questions” section but leave a full evaluation for future work.

The Budget Lab’s State of Tariffs analysis presents economic, fiscal, and distributional estimates of the effect of tariffs on the United States. All these estimates begin with an estimate of the statutory (i.e., in-law) effective tariff rate (ETR) faced by different types of commodities, figures which are estimated using a variety of official government sources, proclamations, and unofficial announcements as tariff policies have evolved. These ETRs — either aggregated into an average ETR or separated by commodity groups — are then fed into either an economic equation (as with our short-run price estimates), macroeconomic models (as with our longer-term economic estimates), or calculations involving the Consumer Expenditure Survey for distributional estimates. The output presented in the State of Tariffs report has evolved over time but generally includes the effect of prices both before and after substitution away from tariffed goods, overall effects on GDP, estimates of the revenue gain from tariffs over a ten-year window, and distributional estimates of the burden of tariffs across households as a function of their after-tax and transfer income.

In this analysis, we focus on consumer prices because they are (a) among the most consequential and direct effects of tariff policy for American households, and (b) observable nearly in real time through PCE and CPI data, allowing us to compare our estimates to outcomes. Other measures, like GDP and employment estimates, are harder to evaluate at this point given lagged data releases and revisions.

On April 2, 2025, we published our analysis of the just-announced suite of country-level tariffs, as well as the previously announced tariffs on Mexico, Canada, and China. Our headline estimate was that the tariff regime in place on that date would increase consumer prices by 2.3%, corresponding to a cost of $3,800 per household per year.

This estimate was derived from a straightforward formula for short-run price effects:

\begin{equation}\Delta P_{sr} = \Delta\tau\times s_{goods}\times s_{import} \times (1+\rho-\delta)\end{equation}

where:

• \(\Delta\tau\) is the change in the ETR from baseline. At the time, we estimated that the overall tariff rate as of April 2nd was approximately 22.4 percent, and the baseline tariff rate as of January 1st was 2.4%, implying \(\Delta\tau=20\%\).
• \(s_{goods}\) is the goods share of PCE, which we estimate to be 33%.
• \(s_{import}\) is the import content share of goods, which we estimated in April of 2025 to be 26%.
• \(\rho\) is the assumed domestic pricing response — the degree to which domestic producers raise prices in response to tariff-driven increases in competing import prices. We use the estimate from Flaaen et al. (2020), which is 50%.
• \(\delta\) is the assumed dollar appreciation offset, or the degree to which the rising value of the dollar due to higher tariffs will offset price increase. As of April 2025, we assumed this value was 15%.

The ETR of 22.4% was the highest U.S. effective tariff rate since 1909. Combined with a 33% goods share, 26% import content, 50% domestic pricing response, and a 15% dollar-appreciation offset, the formula produces a price level increase of 2.3%.

The most important thing to understand about our April 2 estimate is that it reflected a state of the world that never came to pass. This was a point-in-time estimate: it answered the question, “if tariff policy as of April 2 persists unchanged, what would the price effects be?” This "as matters stand at this moment” framework is like that used by the Congressional Budget Office (CBO) for budget scoring — not because we expect current policy to persist per se, but because it provides a consistent baseline for evaluating policy. However, it also did not reflect what became the reality of tariff policy over the course of the year. Tariff rates changed within days of our publication and continued to fluctuate over the course of the year.

Figure 2 shows our daily ETR tracker alongside the ETR we reported in each State of Tariffs edition. Tariff rates peaked briefly after April 2, fell after the Geneva pause, rose again with the 232 expansion and Phase 2 reciprocal tariffs, and then dropped sharply after the Supreme Court struck down IEEPA tariffs in February 2026.

As our first step in analyzing our initial assessment of the effect of tariffs on prices, we want to imagine what our estimate of this parameter would have been had we been fully aware of tariff policies throughout 2025. Across more than twenty editions of the State of Tariffs report published in 2025, our published ETR ranged from 15.6% to 28.0%. To compute a representative average, we assign each edition a weight proportional to our contemporary understanding of long the policy regime it was modeling was in effect. This avoids giving outsized weight to periods with frequent but minor tariff policy changes. The resulting policy-regime-weighted average is 17.5%, substantially below the 22.4% used in our April 2 estimate.

If we re-estimate the price formula using this policy-regime-weighted average ETR instead of the April 2 ETR, we estimate a 1.8% increase in short-run prices. This adjustment accounts for the largest single component of the discrepancy between our April 2, 2025, estimate and the realized price effect of tariffs through early 2026: the tariff regime changed after April 2, and our point-in-time estimate did not (by design) account for subsequent policy evolution.

Even holding the policy regime fixed, we have gotten better at measuring effective tariff rates over the past year. Across 2025 and 2026 we made several substantive improvements to our ETR calculations, including updating our estimates to track the evolving utilization of USMCA exemptions over the summer, improving our modeling of the interactions between USMCA and Section 232 tariffs in October, shifting to using more granular commodity-level data in November, and explicitly modeling the metal content of derivative goods in March 2026. The most recent of these improvements is our Daily Rate Tracker, which provides daily tariff rates for each HS10 commodity and country pair, decomposed by tariff authorities.

While these changes have affected the estimated tariff rate faced by specific countries or products significantly, what turns out to make a bigger difference for the aggregate average statutory ETR are issues surrounding the timing and enactment of policies. In Figure 3, we plot the gap between each State of Tariffs edition’s ETR and the Daily Rate Tracker’s ETR on the corresponding publication date, decomposing that difference into timing and methodology components. In effect, most of the gap comes from the State of Tariffs reports modeling policies that had not yet been enacted (as with July 2025, when we modeled the effect of various tariff policies that were actually implemented in the Fall), were implemented gradually over the course of several HTS releases, or were never actually implemented (as was the case in April 2025). This is an artifact of the fast-moving policy landscape: it was often hard to determine what policies were implemented versus merely announced or threatened as of a certain date. Once these timing issues are accounted for, TBL’s estimates of ETRs have been consistently higher by 1-5 percentage points than our current assessment of the rates. The overestimations of the ETRs stem from a few factors, like our treatment of derivative metal products and assumptions regarding USMCA utilization. It is worth noting that for many of the metrics we produce, the overall ETR is only part of the equation: correctly estimating commodity-level ETRs is an important input into these calculations, and our ETR estimation has also improved along these lines.

We can quantify the impact of this methodological and timing mismatch by estimating the price effects using the daily average ETR from the Daily Tariff Rate Tracker. This was 14.8% from April 2 through December 31, 2025; lower than the policy-regime-weighted average of our published SOT ETRs (17.5%), reflecting the cumulative effect of these improvements. If we re-estimate the price formula using this tracker-based daily average, we get a price effect just over 1.4%.

Beyond measuring tariff rates more accurately, we made several improvements over the course of the year to our economic modeling framework that converts ETRs into price effects and other macroeconomic outcomes.  Figure 4 shows the timeline for these modeling changes. Before today, for the price estimates specifically, the most consequential change came on October 17th when we updated the import content assumption from roughly 26% to 21%, reducing the estimated price impact all else equal.

Today we are releasing our largest update yet for how we estimate pre-substitution price effects. In the past, our price estimate was based on one aggregate equation mapping the overall average ETR shock to the aggregate price impact, mediated by an economy-wide average import content of goods parameter. Now, following Barbiero and Stein (2025), we run commodity-level ETR shocks through an input-output matrix (via BEA). Rather than assume all commodities share the same exposure to import shocks, this approach allows the import content assumption to vary with changes in composition of the tariff regime. For example, our old method could not distinguish between a 1 percentage point overall ETR hike coming from an increase in auto tariffs versus a 1 percentage point overall ETR hike coming from a broad-based tariff; our new method correctly adjusts the import content share parameter to reflect differences across products. This method also makes it easier for us to report the breakdown of price impacts in terms of consumer spending categories instead of commodity categories—another improvement we’re releasing today.

If we re-estimate the price effect using our current methodology (as of March 2026) with the daily average ETR from the previous section, we get an effect just under 1.4%. This number reflects the status of our modeling exercise, incorporating all the methodological improvements we have made since April 2nd of last year. The actual change in the short-run price effect is small (0.04 percentage points), reducing the gap (using a 1 percent price effect as a benchmark) from 40.9 to 37.0 percent.

Even with improved ETRs and updated methodology, there remain questions about the economic assumptions embedded in our model. With a year of data, we can now bound the range of some key parameters. While this is an important task for many of the assumptions in our model, we focus on one for the time being: the passthrough of tariffs to consumers.

On year ago we estimated that the consumer price level would rise by 2.3% in response to the tariff regime as it stood on April 2, 2025. Today, one year later, we can look at the factors which drive the gap between that estimate and the lower retrospective estimate of between 0.5% and 1%:

• Policy changes after April 2 account for the largest share, bringing the estimate down to 1.8%.
• Consistently handling the timing between announced and enacted policy, as well as improvements to our ETR calculations, reduces it further to 1.4%.
• Methodological updates have a modest additional effect.
• The remaining gap can be closed by adjusting our passthrough assumption from 100% to a data-informed range of 40% to 76%, yielding short-run price effects consistent with recent empirical findings.

We still believe our current-policy approach has value. Point-in-time estimates answer the question: if the current tariff regime persists, what are the expected effects? This is the information policymakers, journalists, and the public need to evaluate policy as it stands. It is the same framework CBO uses for budget projections, a framework designed not to be an unconditional forecast of the future but instead a basis on which the public and policymakers can evaluate the relative merits of different policy choices. Our estimates of the economic effects of tariffs are not unconditional forecasts, for the reasons articulated in this article among many others, but we are committed to improving our model as we learn more. Over the past year, we have made substantial improvements to both our ETR calculations and our economic models. We will continue to refine our estimates as new data, new research, and new policy developments warrant. 

The data underlying this analysis is available on Github.