What a Pass-Through Estimate Actually Measures
An exchange-rate pass-through estimate is a measurement of a particular transmission path, not a property of a currency in isolation. This chapter focuses on the distinction between an exchange-rate shock, the price index that responds, the time allowed for adjustment, and the statistical coefficient used to summarize that response. Once those pieces are kept together, the estimates from the global evidence become easier to interpret: they describe how a defined economy and price basket reacted under defined conditions, rather than revealing a constant conversion rate.
From an exchange-rate movement to consumer prices
Start with a nominal effective exchange rate, or NEER. It summarizes the value of a currency against the currencies of its trading partners, giving a broad measure of whether imported goods and services become more or less expensive in local currency. A weaker currency generally raises the local-currency cost of imports; a stronger currency generally lowers it. That first change is economically important, but it is not yet consumer-price inflation. Between the exchange rate and the price paid by a household stand importers, manufacturers, wholesalers, retailers, service providers, taxes, contracts, and domestic demand.
The first route is the imported-inflation channel. An importer buying a foreign product faces a cost in another currency. If the domestic currency weakens and the foreign-currency price does not change, the local-currency cost rises. A firm can pass that increase on to a distributor or customer, absorb it in its margin, alter the product, change suppliers, or wait to see whether the exchange-rate movement persists. The eventual effect on the consumer price depends on how much of the imported content is in the product and how much bargaining power each firm has along the chain.
The second route is a derived-demand channel. An exchange-rate movement changes the demand for domestic output and for factors such as labor and intermediate inputs. In a small open economy, an appreciation can reduce demand for domestic production and narrow a positive output gap; weaker demand can then ease pressure on wages, hiring, investment, and prices. A depreciation may work in the opposite direction if it supports demand for domestic production. This route is not a duplicate of the imported-cost route. One works through the cost of what firms buy from abroad; the other works through the demand for what the domestic economy supplies.
These routes meet the consumer price index at different distances and speeds. Import prices are close to the initial cost change. Producer prices may reflect imported inputs before retail prices do. Headline CPI also includes domestic services, housing-related items, taxes, and other components whose costs are not determined directly by the exchange rate. Core CPI may exclude volatile items. Goods can respond earlier than services because goods often contain tradable inputs, while services rely more heavily on local labor, capacity, and contracts. An estimate for import prices over a quarter is therefore not interchangeable with an estimate for headline CPI over a year.
It is useful to think of three filters between the exchange-rate movement and the published number. The first is firm-level pricing: how much of the cost change reaches the next price, and how much is absorbed in a margin or delayed. The second is aggregation: which prices receive large weights in the index, and whether a movement concentrated in tradable goods is diluted by services. The third is dynamics: whether the adjustment appears immediately, unfolds as inventories are replenished, or is revised when contracts are renegotiated. A pass-through coefficient is the result after all three filters have operated. It is not a transparent window into any one firm’s markup decision.
Consider a deliberately simple hypothetical. A currency movement raises the local-currency cost of an imported input, but a retailer has inventory purchased before the movement and a contract that fixes its selling price for several months. The immediate consumer-price response may be small. When the inventory runs out or the contract is renewed, the response may become larger. If domestic demand weakens at the same time, the retailer may instead accept a smaller margin. The one-year outcome could then be modest even though the original import cost changed materially. The example does not predict a particular coefficient; it shows why the timing and endpoint are part of the economic claim.
The sequence also explains why “pass-through” is not the same as “the exchange rate caused every price change.” Firms are responding to the exchange rate inside an environment that includes inflation, the output gap, global commodity prices, financial conditions, expectations, and policy decisions. The statistical task is to estimate the exchange-rate component while allowing for other influences. Even a well-specified estimate remains a response of an index that contains many prices, not a direct observation of each firm's decision.
The time horizon is part of the claim. If a currency moves today, some prices may adjust in the same quarter, others in the following quarter, and still others only when contracts are renewed or inventories are replaced. A one-year pass-through can therefore be larger than a same-quarter response without implying that the later increase was instantaneous. Conversely, a small one-year estimate does not prove that the exchange rate never matters; it may mean that the movement was offset, delayed, or measured at an endpoint dominated by less-tradable prices.
An estimate is also a choice about what counts as the exchange-rate movement. A broad NEER can move because several trading-partner currencies move at once, while a bilateral rate can highlight a different exposure. A market-driven movement and a policy announcement may have different information content for firms. If the movement is expected to reverse, price setters may wait; if it signals a persistent change in costs, they may adjust sooner. This is why two studies can use the same phrase—exchange-rate pass-through—while answering different questions. The label is not enough; the shock definition must travel with the result.
There is a further interpretive issue: the movement being measured may not be an exogenous shock. In Singapore, for example, the exchange rate is the central monetary-policy instrument. A monetary authority can strengthen the currency because it expects inflationary pressure to rise. If an analyst simply compares the currency with subsequent inflation, the observed relationship mixes the effect of the currency movement with the reason policymakers made that movement. The direction of the bias cannot be understood from the coefficient alone. The exchange-rate change must be defined as carefully as the price index and horizon.
How the BIS turns dynamics into a comparable estimate
The Bank for International Settlements study makes these choices explicit in a broad cross-country setting. It uses quarterly consumer-price inflation, nominal effective exchange rates, and output gaps for economies with enough observations to estimate relationships within each decade. The study retains 98 economies, and its coverage grows from 72 countries in the 1980s to 98 in the 2010s. That expanding coverage is useful for learning about a wide range of economies, but it also means that decade comparisons should not be read as if the set of countries were identical in every period.
The first-stage model follows a mean-group approach. In plain terms, it estimates the relationship for each country rather than imposing one common exchange-rate coefficient on the entire panel. Each country has its own response to current and lagged NEER movements, as well as its own dynamics for inflation and the output gap. A country-specific intercept also absorbs fixed differences, including differences in inflation targets as the paper notes. The model includes lagged inflation to capture persistence, the output gap to account for domestic price pressure, and the exchange-rate movement in the current quarter and four subsequent quarterly positions.
Why not simply compare inflation before and after a currency movement? Because the before-and-after change would combine the exchange rate with everything else that changed in the same period. Inflation can persist from earlier quarters; an economy can be overheating or underusing capacity; oil prices can move; financial stress can spread across borders; and a global crisis can affect both currencies and prices. The BIS specification does not solve every identification problem, but its lagged inflation, output-gap, oil-price, volatility, and crisis controls make the comparison more disciplined. They help distinguish a dynamic exchange-rate relationship from an undifferentiated change in the price index.
That timing structure is what turns a movement into a dynamic response. The contemporaneous coefficient asks how much of the relationship appears in the current quarter, while the lagged coefficients allow the adjustment to continue or reverse. The paper also reports a version controlling for changes in Brent crude prices, the CBOE volatility index, and the global financial crisis of 2008–09. These controls do not make the estimate a perfect experiment. They are a way to avoid attributing every international disturbance to the exchange rate when oil prices, financial stress, or a global crisis are moving at the same time.
The resulting coefficients can be added over a one-year horizon. This sum is the study's one-year pass-through measure: it captures the estimated response in the current quarter plus the four quarterly lags. It is a compact summary of a dynamic path. The full sample's average one-year estimate is 13.4 percent, and the paper reports that most of the effect occurs in the same quarter or the quarter immediately after the exchange-rate movement. The number is thus best read as a cumulative response under the paper's exchange-rate convention, not as a claim that 13.4 percent of a currency movement appears immediately in every consumer price.
Aggregation creates a useful trade-off. Summing the lags gives readers a common horizon and makes country-period estimates easier to compare. But the sum can hide important timing differences. A response concentrated in the current quarter may matter differently for monetary policy and household expectations than an equal response spread across four quarters. A later coefficient can also offset an earlier one, leaving a moderate cumulative number that does not mean the intervening price movements were negligible. The one-year figure is therefore an index of the path, not a replacement for the path.
Speed and magnitude answer different questions. Two economies could have the same one-year estimate while one adjusts rapidly and the other adjusts gradually. They could also have similar same-quarter responses but different later movements because firms revise prices, inventories change, or domestic demand evolves. A coefficient summed across lags is valuable because it gives a common horizon, yet the sum should not erase the timing information that generated it.
The mean-group design matters for a second reason. Suppose the true responses were low in one group of countries and high in another. A pooled model that forced the same slope on everyone would return a central estimate but would treat the differences as unexplained noise. The BIS procedure instead produces country-period estimates that can later be related to inflation, volatility, economy size, product composition, and policy conditions. Heterogeneity is not a nuisance added after the main result; it is the object that the second stage is designed to explain.
The country-period unit also sets a boundary on interpretation. The second stage does not ask why one shipment was repriced while another was not. It asks why the estimated relationship for a country in a particular decade tends to differ from the relationship elsewhere. That is an appropriate scale for studying macroeconomic and structural conditions, but it cannot identify every firm-level channel. A country-decade estimate can be informative about an environment without describing the behavior of every sector within it.
This design also clarifies what an estimate does not contain. It does not observe the counterfactual price path that would have occurred if the exchange rate had not moved. It does not identify a single markup rule for every firm. It does not tell us that the included controls have removed every confounding influence. And because CPI is an aggregate index, it does not reveal whether a measured response came mainly from imported goods, domestic services, or another component without a separate decomposition.
The sign requires the same discipline. The raw coefficient in a regression depends on how the NEER is defined—for example, which direction is labeled an increase—and on whether the analyst is discussing an appreciation or a depreciation. The economically useful statement is not simply “the coefficient is positive” or “the coefficient is negative.” It is: under this convention, over this horizon, a specified exchange-rate movement is associated with a specified cumulative change in the chosen price index. When comparing studies, preserving that convention prevents a numerical sign from being mistaken for a different economic conclusion.
The estimate also inherits the sample's boundaries. The BIS study uses an output gap that begins as annual data and is linearly interpolated to quarterly frequency. It requires a minimum amount of quarterly data within a decade, and it trims the most extreme estimated pass-through observations before examining their drivers. Those choices make the analysis more stable and interpretable, but they define which economies and periods contribute to the result. A number estimated from a balanced sample, a country-specific time series, or an identified policy shock would answer a different question.
The practical lesson is simple but demanding: never detach a pass-through number from its measurement frame. Identify the exchange-rate variable and its sign convention. Identify the price endpoint and whether it is an import, producer, headline, core, goods, or services measure. Identify the horizon and whether the number is an immediate response or a cumulative sum of lags. Finally, ask how the exchange-rate movement was generated and which other forces the model accounts for. Once that frame is in place, the global averages become useful evidence rather than false constants—and the important question becomes why those country-period responses differ in the first place.