In short: Absorb or pass through is the wrong unit of decision, because tariff cost pass-through pricing has a different answer by item, by channel and by customer. The calculation rests on a landed cost per item that many businesses cannot currently compute to the accuracy the decision needs. A tariff response differs from an ordinary price move in three ways: the change is large, it may reverse, and competitors face it at the same time. A pass-through decision changes demand and therefore the supply plan, so taking the two in separate meetings is how an announced increase triggers a forward buy the network cannot serve.
Tariff volatility has become the dominant regulatory concern for trade professionals. In 2026 surveys, seventy-two percent named it the most impactful regulatory change, up from forty-one percent the year before, and more than three quarters of respondents said they expect the current approach to persist for at least four years rather than being a negotiating position.
The behavioural response has shifted with it. Thirty-nine percent now report absorbing or considering absorbing tariff costs rather than passing them on, up from thirteen percent a year earlier. Among smaller businesses the pattern is the opposite, with eighty-two percent passing costs through, overwhelmingly by direct price increases.
Those two facts together describe a market where the same cost shock is being handled in opposite ways by different firms, which means the decision is not obvious and is worth doing properly.
The economics literature from the previous round of tariffs is worth knowing before you start, because it says something specific about where the cost lands. Amiti, Redding and Weinstein looked at the 2018 US tariffs in the Journal of Economic Perspectives in 2019 and found essentially complete pass-through into duty-inclusive import prices, meaning foreign exporters did not absorb the tariff by cutting their prices. Cavallo, Gopinath, Neiman and Tang then traced the same tariffs further down the chain in the American Economic Review: Insights in 2021 and found that pass-through at the border was close to complete while pass-through at the retail shelf was considerably smaller.
The gap between those two findings is where this decision lives. Somebody in the chain between the port and the shelf absorbed a substantial share of that cost, and the empirical answer from the last cycle is that it was largely the importer and the retailer rather than the exporter or the consumer. If your planning assumption is that the supplier will share the pain or the shopper will carry it, the evidence from the most recent large episode runs against you on both counts.
Absorb or pass through is the wrong framing
Treating this as a single binary produces a bad answer, because the right response differs by item, by channel and by customer, and averaging across them loses the decision.
The useful framing is a portfolio one. For each product, the question is what share of the cost increase to pass through, and the answer depends on three things you can estimate.
Own-price response. How much volume you lose per point of price. Where response is weak, pass-through is cheap. Where it is strong, a full pass-through loses more margin through volume than it recovers through price.
Competitive position. If competitors face the same cost increase, pass-through is far safer than if the increase is specific to your sourcing. This is knowable: tariff schedules are public, and where your competitors source from is usually inferable.
Channel constraints. Some channels have contractual price protection, notice periods or annual negotiation windows that make a price move impossible this quarter regardless of what the analysis says. Those constraints bind first.
Run those three across the portfolio and what falls out is a mixed answer: full pass-through on some lines, partial on others, absorption on a small set where the position is weak, and in several cases a change to pack or format rather than a price move at all.
An item's worth of arithmetic makes the shape of the answer clear. Take a product landing at 9.35 and selling at 14.00, so a contribution margin of 4.65, moving 20,000 units and producing 93,000 of margin. Duty goes from five percent to twenty-five percent on a customs value of 8.00, so the duty line goes from 0.40 to 2.00 and landed cost becomes 10.95. Full pass-through means a price of 15.60, an eleven point four percent increase, which restores the unit margin exactly.
Now price three options at an own elasticity of minus one point five. Full pass-through drops volume seventeen percent to about 16,570 units at 4.65, giving 77,000 of margin. Half pass-through puts the price at 14.80, unit margin 3.85, volume down about nine percent to 18,290, giving 70,400. Absorbing it entirely holds 20,000 units at 3.05, giving 61,000. Full pass-through wins, and all three are well below the 93,000 you started with, which is the part nobody wants to say out loud in the meeting.
Rerun the same item at an elasticity of minus three and the ranking changes. Full pass-through now gives about 61,100, half pass-through gives about 63,800, and absorption still gives 61,000. Partial becomes the best of a poor set.
Two things follow. The optimum moves from full toward partial as elasticity rises, so the portfolio answer really does have to be computed item by item. And the differences between the options are small relative to the total loss, which means precision in the elasticity estimate is worth much less than knowing which side of about minus two and a half the item sits on. For this item the two options cross over at around minus two point six. That is a considerably easier question to answer under time pressure, and it is the one to put to the commercial team.
Landed cost has to be right first
A surprising amount of the difficulty here is that many businesses cannot compute landed cost per item accurately.
Landed cost is the purchase price, plus freight, plus duty at the applicable rate for the applicable classification and origin, plus insurance, plus port and handling charges, plus any brokerage. Each of those varies, and the duty component in particular depends on tariff classification and rules of origin that are more contestable than most people assume.
Two things that catch businesses out.
Classification. The tariff code assigned to a product determines the rate, and codes are frequently assigned once at product setup and never reviewed. A misclassified product can be paying a materially wrong rate in either direction, and correcting it is one of the more direct cost improvements available.
Origin rules. Where a product counts as originating from is a legal determination based on where substantial transformation occurred rather than simply where it shipped from. For assembled goods with components from several countries this is genuinely complex, and it is the lever that makes some supply chain reconfiguration worthwhile.
Getting landed cost right per item is the prerequisite. A pass-through decision built on an average duty rate applied across a portfolio will be wrong item by item in ways that cancel in the aggregate and matter individually.
The example above shows why the classification question is worth real effort. On a customs value of 8.00, the difference between a five percent code and a twenty-five percent code is 1.60 per unit, which on that item is thirty-four percent of the contribution margin. A single product sitting in the wrong code at twenty thousand units a year is a thirty-two thousand error, running every year, with no line item anywhere that names it.
The diagnostic is a phone call. Ask your customs broker for a duty-paid report for the last twelve months, broken out by tariff code and by product, sorted by duty paid. Most businesses have never requested that file and most brokers can produce it in a day. The top twenty lines will normally carry the great majority of the duty, which tells you exactly where a classification and origin review is worth the professional fees and where it is not.
The dynamics that make this different from a normal price move
Three features distinguish a tariff response from an ordinary pricing decision.
The change may reverse. Trade policy moves in both directions, and a price increase is much harder to unwind than to implement. This asymmetry argues for mechanisms that are easier to reverse than a list price change: a temporary surcharge, a reduction in promotional depth, a change in pack size, or a shift in the trade terms rather than the price.
Timing interacts with inventory. Goods already landed were bought at the old duty rate, and goods on the water may or may not be. A price increase implemented while selling old-cost inventory is a margin windfall; one implemented too late means selling new-cost inventory at old prices. Aligning the price change to the cost cohort actually being sold requires knowing your inventory ageing, which is a planning input rather than a commercial one.
Everyone is doing it at once. When a tariff applies to a category broadly, competitor pass-through becomes the dominant factor in whether yours sticks. Moving first carries risk; moving last carries margin cost. This is a judgement rather than a calculation, and the calculation should tell you how much each week of delay costs so the judgement is informed.
The supply consequence belongs in the same decision
A pass-through decision changes demand, and changed demand changes the supply plan. Making the two decisions in separate meetings is how commercial teams announce a price increase that triggers a forward buy the network cannot serve.
Two effects to model before committing.
A price increase announced with notice produces a pull-forward as customers buy ahead. That spike is real, it is temporary, and it will be followed by a trough. Planning production against the spike is a well-known way to end up with a quarter of unwanted stock.
The arithmetic is worth spelling out because it is larger than it feels. A distributor who normally holds four weeks of cover and is given thirty days' notice of a ten percent increase has an obvious trade: buy ten weeks instead of four and bank the difference. If a meaningful share of the customer base does that, the notice month runs at something like two and a half times normal, and the two months after it run at a fraction of normal while the extra cover works off. Nothing has been sold that would not have been sold anyway. All that happened is that a quarter's volume moved.
The second-order failure is the one that catches people twelve months later. That spike goes into the demand history as a very large month with no promotional flag against it, and next year's statistical model reads it as a seasonal peak and forecasts one. Nobody connects the phantom peak to the tariff announcement from the previous year, because the person running the forecast is not the person who signed the price letter. Flag the affected periods in the history at the time you announce, with a note saying what caused them, and treat the pull-forward and the trough as a matched pair when you clean the series. Doing that costs ten minutes in the month it happens and is close to unrecoverable a year later.
A price increase that is not matched by competitors produces a volume decline that has to flow into the supply plan, or you will build to a demand curve that no longer applies.
The commercial and supply handshake before the announcement is the mechanism, and the check is straightforward: what does the network do under the expected volume path, and is that feasible.
What to prepare in advance
Because these changes arrive with little notice, the response quality is largely determined by what exists beforehand.
A landed cost model that can be reparameterised in hours rather than weeks, with duty rates as inputs rather than embedded assumptions.
Elasticity estimates by item and channel, even rough ones, so that the pass-through decision starts from evidence rather than from a debate.
A pre-agreed decision framework specifying who decides, what evidence is required, and what the approval path is for each size of move. When the change lands, the constraint is usually the decision process rather than the analysis.
The limit
Elasticity estimates for a shock of a size you have never experienced are extrapolations. If your history contains price moves of two or three percent and you are contemplating twelve, the estimate is being used well outside its range, and the honest treatment is a wide band with the recognition that consumer response to a large increase is often non-linear in ways a small-move elasticity cannot predict.
There is also a limit on how far pricing can be the answer. Where a tariff makes a sourcing arrangement structurally uncompetitive, the durable response is a supply chain change rather than a price change, and pricing analysis buys time rather than solving it. Knowing which situation you are in matters more than optimising the price, and the test is whether the increase is large enough that a competitor with a different sourcing footprint can undercut you sustainably.
Start with the landed cost model per item. Every decision described here depends on it, and most businesses discover the model needs work before the first tariff conversation is over.