When the Average Gets Cheaper by Changing Suppliers
The question in this chapter is how the 2025 U.S. tariffs changed the mix of foreign suppliers inside an imported product, and why that can make an average price look lower even when individual exporters barely change their quotes. The reader should come away able to separate repricing by continuing suppliers from entry and exit, and to judge why a nominally cheaper mix is not automatically a better deal for consumers or firms using imported inputs.
This is the upstream counterpart to the household price question. A product sold in the United States is not always a single, stable object in the trade data. It may be a category assembled from several source countries, each supplying a variety with its own price and appeal. When tariffs alter which varieties are sold or gain market share, the average can move because the composition moved. The average is then reporting both prices and the population of suppliers behind those prices.
A product is an average of varieties
The IMF study works at two levels. A “variety” is an exporter–HS10 product combination: for example, one detailed product supplied by one country. A “product” is the HS10 category after the varieties from different exporters have been aggregated. That distinction creates two different questions. Did a given exporter change the price of the variety it continued to sell? And did the product’s source mix change so that one set of varieties counted more heavily than another?
The first question is the intensive margin. It concerns varieties that appear in both the earlier and later comparison periods. If an exporter keeps selling the product and lowers its duty-exclusive price, that is repricing by a continuing supplier. The second question is the extensive margin. It concerns the boundary of the market: which varieties enter, which exit, and which are no longer present in the product’s average. A product can experience little intensive-margin repricing and a large extensive-margin change at the same time.
That distinction is not just a classification choice. An average price is constructed from the prices of the varieties that are observed and the shares those varieties receive. If a high-priced source exits, the average can decline even if every continuing source keeps its price unchanged. If a lower-priced source enters or gains share, the average can decline for the same reason. The product-level result is therefore sensitive to the identity and weight of the suppliers, not only to their posted prices.
The IMF’s variety-level estimates provide the starting condition for this mechanism. In the baseline 2025 results, tariff changes have essentially no effect on duty-exclusive exporter unit values. Duty-inclusive import prices, which add the collected duty, rise one-for-one with the effective tariff. At the same time, import values and quantities fall. An eight-percentage-point increase in the effective tariff, similar to the average increase from February through December 2025, implies an average import decline of about 3.6 percent.
That combination means that the tariff did not generally produce a broad round of exporter price cuts that neutralized the duty. Instead, some of the adjustment occurred in trade flows. Varieties facing higher tariffs lost sales relative to other varieties of the same product. The baseline evidence also shows that higher statutory tariffs increase the probability that an existing variety exits the U.S. market. The estimated rise is about 1.4 percentage points when the average statutory rate moves from roughly 2 percent to 15.5 percent, against an average exit probability of about 26 percent.
Entry has a more subtle interpretation. Higher statutory tariffs reduce the probability that a variety at risk of entering appears in the U.S. market: the same policy change lowers the average entry probability by about 0.3 percentage points, from a baseline near 1.6 percent. But among the varieties that do enter in the product-level price decomposition, the entering varieties are relatively low-priced in products facing larger tariff increases. Thus “entry” can be a smaller doorway overall while still having a downward effect on the average price of the product represented by the varieties that pass through it. The number of entrants and the price position of entrants are different margins.
The product-level estimates make the compositional effect visible. An approximately eight-percentage-point increase in average effective tariffs is associated with about a 5 percent decline in before-tariff import prices at the HS10 product level. Around 65 percent of the decline is attributed to the extensive margin. Within that extensive contribution, the paper attributes roughly 65 percent to lower-priced entering varieties and 35 percent to higher-priced varieties exiting. The remaining adjustment comes through intensive-margin components, including changes among continuing varieties and their import shares.
This does not mean that tariffs made imported products intrinsically cheaper. It means that the product-level index increasingly represented a different selection of sources. The index is answering, “What was the average price of the varieties being imported in this product?” It is not answering, “What would the same fixed basket of varieties have cost absent the tariff?” Those questions converge only when composition is stable.
The mechanism can be pictured as a small system rather than as a single price line:
The top half describes the source-composition channel. The bottom half asks what the composition means. These are related but not identical questions. A low-price source may be more efficient, allowing the importer to cushion some of the tariff. Or it may supply a less appealing variety, so the nominal price reduction partly reflects a quality or preference loss. The data can show the correlation between tariff exposure and the resulting mix; interpreting its welfare meaning requires a further adjustment.
The cheaper mix may not be a better deal
To examine that ambiguity, the IMF authors estimate “appeal” using a CES demand framework and pre-tariff data from 2023–24. Appeal is a demand shifter: a variety that attracts a relatively large import share despite a higher unit price is assigned greater appeal in the model. The measure combines differences that might be described as quality with differences in consumer or importer preferences. It is not a physical inspection score and should not be treated as a direct observation of durability, performance, or safety.
The authors then construct two kinds of product-level price index. The unadjusted index aggregates unit prices without accounting for the fact that varieties may have different appeal. The appeal-adjusted index weights prices in a way intended to compare the cost of obtaining utility-equivalent varieties. Both can be calculated with duty-exclusive prices, which describe the pre-duty cost at the border, or duty-inclusive prices, which are more relevant to the buyer’s cost after tariffs.
The 2025 estimates show why the distinction matters. For duty-exclusive prices, the tariff coefficient is about −0.642 in the unadjusted index but about −0.357 in the appeal-adjusted index. Both are negative, so the evidence still indicates a decline in quality-adjusted before-tariff product prices. But the decline is much smaller after appeal is held in view. The difference implies that roughly half of the nominal reduction is associated with movement toward lower-appeal varieties rather than with a pure switch to cheaper sources of comparable appeal.
For duty-inclusive prices, the contrast is sharper. The unadjusted coefficient is small and statistically insignificant, about 0.059. On that measure, the product-level price appears not to respond meaningfully to the tariff once the import mix is allowed to change. The appeal-adjusted coefficient is positive and significant, about 0.375. Once the index accounts for differences in appeal, the evidence points to a rise in the welfare-relevant import price, although less than one-for-one with the average tariff change.
The result changes the meaning of “the tariff lowered import prices.” At face value, a lower unadjusted average looks like a benefit: buyers appear to have found cheaper sources. After appeal adjustment, part of that apparent benefit becomes a cost of obtaining a less preferred mix. The importer may have reduced its cash outlay per unit while receiving a product with less demand appeal, or while accepting a source that consumers value less. For a firm, the analogous issue is the quality of an imported input: a lower invoice price need not mean a lower cost per unit of productive service.
The paper does not settle how much of the reallocation is efficiency and how much is quality loss. It offers two plausible mechanisms. Large tariff increases may make it worthwhile for importers to incur the fixed costs of searching for alternative suppliers, including more efficient low-cost sources. But the same reallocation can push trade toward varieties with lower appeal. Search and matching costs, source capability, and buyer preferences can operate together. The appeal estimates establish that quality or preference composition matters; they do not identify a single cause for every supplier change.
This is also where the household evidence must be kept at arm’s length. The import study can show that the mix of varieties entering the United States changed and that the appeal-adjusted price response differs from the nominal response. It cannot show that a particular household bought a particular lower-appeal product. The household study can show trade-down within tariff-affected categories, where spending falls while quantities hold, but it does not identify the foreign supplier behind each retail purchase. The two findings are conceptually compatible, but they are not the same observation.
The 2018–19 tariff episode provides a useful test of whether this mechanism is unique to the broad 2025 tariff changes. In the earlier episode, the negative product-level relationship between tariffs and duty-exclusive import prices is stronger and is driven almost entirely by the extensive margin. Lower-priced varieties enter and higher-priced varieties exit in products facing larger tariff increases. The appeal-adjusted duty-exclusive coefficient is about −0.421, compared with −1.024 for the unadjusted index. For duty-inclusive prices, the appeal-adjusted coefficient is positive at about 0.274, while the unadjusted coefficient is negative at about −0.267.
The recurrence strengthens the mechanism: source composition and appeal can materially dampen measured tariff pass-through across more than one U.S. tariff episode. It does not make the episodes identical. The earlier tariffs focused mainly on China, used a different policy environment and comparison window, and relied on a different source for statutory tariff data. The comparison is therefore evidence that the mechanism can recur, not proof that the 2018–19 magnitudes forecast the 2025 household response.
There are further reasons to resist reading the monthly estimates as a permanent price transformation. The IMF results remain qualitatively similar with quarterly data, but the coefficients are about half as large, suggesting that some monthly movement reflects lumpy shipments or high-frequency changes in unit values. The appeal measure is estimated from the pre-period and treated as time-invariant. The analysis is also based on import unit values, which can conceal firm-level pricing and product variation. These choices do not erase the central result, but they define its scope.
The important conclusion is therefore conditional but substantial. Tariffs can lower an unadjusted product-level import average by changing who supplies the product. That composition effect can cushion the visible price impact at the border and downstream. Yet if the new mix has lower appeal, the nominal cushion overstates the welfare gain. A product can become cheaper in the accounting sense while becoming more costly in utility-equivalent terms. This is why a complete tariff analysis must track not only the price of a continuing supplier, but also the suppliers that disappear, the varieties that appear, and the attributes hidden by the average.
With that upstream mechanism in view, the household question becomes more precise. When the price and source mix behind a category change, households still choose whether to remain in the category, how many units to buy, and which variety to accept. The next step is to examine why those choices produced a spending contraction much larger than the measured retail-price increase.