I. Introduction

The choice of invoicing currency in international trade has been recognized as a key factor determining the transmission channels of external shocks, as invoicing currencies can significantly influence the terms of trade and the expenditure switching channel due to exchange rate fluctuations. Under the traditional Mundell-Fleming model (Mundell, 1963; Fleming, 1962), assuming producer currency pricing (PCP), it is known that depreciation of the domestic currency improves the trade balance. In contrast, as suggested by Devereux and Engel (2003) and Engel (2011), if international trade is priced in the importer’s currency (local currency pricing: LCP), exchange rate fluctuations may fail to produce significant adjustments. Alternatively, under dominant currency pricing (DCP), fluctuations in the dollar’s value have been found to affect international trade even when bilateral exchange rates with trading partners remain unchanged (Gopinath et al., 2020). Therefore, understanding invoicing currencies in international trade is essential for accurately assessing the impact of exchange rates and offers macroeconomic implications regarding the transmission channels of external shocks and optimal policy responses.

With this motivation, this paper studies invoicing-currency patterns in Korean trade and examines (i) what determines firms’ invoicing-currency choices and (ii) how those choices shape exchange-rate effects on export prices and quantities. Korea provides a particularly informative laboratory: it is a trade-dependent small open economy whose currency is not used as a vehicle currency. Unlike existing studies using highly granular transaction-level data — which have focused on vehicle-currency issuers such as the United States or the Euro area — this setting helps to distinguish DCP cleanly from other pricing choices (PCP or LCP) that are difficult to separate in data from dominant-currency issuers.

First, we document stylized facts using transaction-level customs data. Korea’s trade is concentrated in a relatively narrow set of invoicing currencies relative to its number of trading partner countries. The U.S. dollar accounted for the majority of export and import value over 2010–2022, while the share amount for won remains small but gradually rising. In contrast, in terms of the number of trade types (i.e., distinct product–destination pair), the won’s share is relatively large — suggesting that won invoicing is more prevalent in smaller-value transactions but spans many products and partner countries. These descriptive patterns are consistent with the dominance of the U.S. dollar in international trade while also highlighting meaningful heterogeneity by currency and by the unit of observation (value versus trade types). Our findings shed new light on invoicing behavior in a small open economy without vehicle-currency status, revealing that the domestic currency plays a more important role than aggregate value-weighted data would suggest. Relying solely on aggregate invoicing statistics may understate the extent to which Korean firms invoice in won, particularly among those operating at smaller transaction scales.

We then examine the determinants of the invoicing-currency choice. Following Engel (2006) and Amiti et al. (2022), we show that factors governing the degree of exchange-rate pass-through also shape invoicing decisions. Specifically, we find that (i) products with a higher share of foreign-currency intermediate inputs are less likely to be invoiced in won and that (ii) products with a higher share of dollar denominated intermediate inputs are more likely to be invoiced in dollars. A higher market share in a destination is associated with a lower likelihood of dollar pricing, consistent with higher markup elasticities making non-dollar pricing (e.g., local-currency pricing) more attractive. We investigate as well how the currency composition of firm liabilities — a factor known to influence exchange-rate pass-through (Kim and Lee, 2024; Kim et al., 2025) but largely overlooked in the invoicing literature — affects the choice of currency. We find that firms with greater foreign-currency liabilities are more likely to invoice exports in foreign currency.

Consistent with Crowley, Han, and Son (2020), we find that currency choices are strongly path-dependent — past invoicing choices predict the current invoicing status. We also document the evidence of strategic complementarity, with firms tending to align their invoicing currency with competitors within narrowly defined markets, as in Amiti et al. (2022). We further show that the strength of these complementarities varies systematically with the degree of demand elasticity, a dimension not previously documented, with firms in more elastic markets exhibiting stronger incentives to coordinate. This finding highlights how buyer-side characteristics, most notably demand elasticity, play an independent role in shaping currency choices.

In line with Goldberg and Tille (2016), we find that greater volatility between the destination currency and the Korean won is associated with a greater share of dollar invoicing. At the same time, we additionally find that greater volatility between the U.S. dollar and the destination currency, as well as a larger foreign exchange market size of the destination currency relative to the dollar, reduces the share of dollar invoicing. Goldberg and Tille (2016) argue that such shifts in dollar share are driven by adjustments in PCP; however, we find no significant adjustment in won invoicing in response to these factors. These findings suggest that the crowding-out of PCP emphasized in the existing literature is less applicable to Korea, a small open economy whose currency does not serve as a vehicle currency. They also imply, albeit cautiously, that efforts to expand the foreign exchange market size or stabilize exchange rates may not by themselves lead to a substantial increase in won invoicing.

We then show that these invoicing patterns have clear macroeconomic consequences. Exchange-rate pass-through into export prices differs significantly across invoicing currencies at short horizons: pricing in won versus non-won, and pricing in dollars versus non-dollars. The estimated differentials decline gradually, indicating that the importance of the invoicing currency diminishes over time as firms gradually adjust prices.

We also confirm that exchange-rate-driven export price changes translate into real allocation. Specifically, we instrument export-price changes with exchange-rate movements in a 2SLS framework following Amiti et al. (2022) and find that export quantities respond negatively to price changes. The implied price elasticity increases in absolute value over time, consistent with a J-curve-type dynamic in which quantities react to price changes with a lag. We subsequently document substantial heterogeneity in elasticities across products, suggesting that the composition of Korea’s export basket can affect aggregate quantity responses.

We find that the invoicing structure of Korean exports has significant macroeconomic implications during periods of dollar strength. A counterfactual analysis shows that if dollar-invoiced exports had instead been invoiced in the importer’s local currency, the export volume growth in October of 2022 would have been higher by 3.05 percentage points. Under a hypothetical scenario of universal won invoicing, the corresponding increase rises to 5.24 percentage points. These findings suggest that the prevailing dollar-invoicing convention acted as a meaningful constraint on Korea’s export responsiveness during a period of pronounced dollar strength.

This paper provides a comprehensive analysis of the determinants of invoicing currency and how the invoicing currency shapes exchange-rate transmission in a small open economy. By documenting invoicing-currency patterns in Korean trade using transaction-level data, identifying the key determinants of currency choices, and quantifying the implications of these choices for export prices and quantities, the paper clarifies how exchange-rate movements propagate through trade. The insights are highly relevant for policymakers in small open economies that do not issue vehicle currencies and nevertheless face pervasive dominant-currency pricing, which can materially shape the short-run effects of external shocks on trade outcomes.

A growing body of research examines the determinants of invoicing currencies. Ito and Chinn (2013) show that cross-country differences in macroeconomic conditions between exporters and importers affect the currency of invoicing, with greater financial market openness found to be associated with a higher share of transactions invoiced in the domestic currency. Using Canadian import customs data, Goldberg and Tille (2016) study both macro- and micro-level drivers and find that transaction size is an important determinant; they also document important roles for exchange rate volatility, the exchange rate regime, and the foreign exchange market depth. Because they use Canadian import data, their framework is not confined to exporters located in small open economies. Drawing on firm-level Belgian customs data, Amiti et al. (2022) demonstrate that firms’ invoicing-currency choices are shaped by the share of imported intermediate inputs, the market share, and by their competitors’ invoicing choices. Finally, using UK exporter data, Crowley, Han, and Son (2020) find that firms’ prior experience significantly influences their invoicing currency decisions. The differential effects of exchange rates based on invoicing currency have also been examined in various studies (Chung, 2016; Devereux, Dong, and Tomlin, 2017; Chen, Chung, and Novy, 2019; Auer and Lein, 2021; Amiti et al., 2022; Corsetti, Crowley, and Han, 2022). Most existing studies focus on advanced economies with deep foreign exchange and financial markets and, in many cases, on issuers of dominant vehicle currencies. As a result, their implications may not readily generalize to small open economies that do not issue a vehicle currency; in such settings it is often difficult cleanly to distinguish PCP from DCP. This study contributes to the literature by examining invoicing behavior in a small open economy where the roles of PCP and DCP are clearly distinguishable.1

The remainder of the paper is organized as follows. Section II describes the distinctive features of Korea’s export and import invoicing currencies. Section III examines the determinants of invoicing-currency choices in Korean exports. Section IV studies how exchange-rate pass-through varies by invoicing currency for Korean exports and the corresponding implications for real allocations. Section V concludes the paper.

II. Data Description

We use transaction-level customs data from the Korea Customs Service, which reports import and export transactions by item, partner country, and currency. Products are classified at the 10-digit HSK level, and the data range from 2010 to 2022, covering all of Korea's trading partners. In the data description section, we employ data aggregated at an annual frequency, while the empirical analysis in Section 3 onward relies on monthly frequency data.

From 2010 to 2022, Korea exported goods to an average of 237 countries per year and imported goods from an average of 240 countries per year. However, as shown in Table 1, the range of currencies used in Korea’s trade is relatively narrow. In particular, only 46 currencies were used for export transactions, and 40 for import transactions. Some trading partners share a common currency (e.g., the Euro) or use the U.S. dollar as legal tender. Even after accounting for these cases, the number of currencies involved in Korean trade remains far smaller than the number of trading partners. This pattern indicates that Korea’s international trade is heavily concentrated in a limited set of currencies.

TABLE 1

NUMBER OF TRADING PARTNERS AND INVOICING CURRENCIES

Note: All figures were initially computed for each year and then averaged over the 2010–2022 period.

The U.S. dollar has consistently accounted for the majority of Korea’s exports and imports by value. The pattern remains similar even when excluding countries that use the U.S. dollar as legal tender, indicating that the dollar’s dominance is not driven solely by such trade partners. In particular, from 2010 to 2022, the share of dollar-invoiced trade value remained relatively stable – around 75% for exports and 83% for imports. Overall, Korea’s invoicing patterns align closely with the Dominant Currency Paradigm (Gopinath et al., 2020).

In contrast, the currencies of Korea’s trading partners accounted for a much smaller share of Korea’s trade than the U.S. dollar, remaining at roughly 10% in both exports and imports over the period of 2010-2022. The Korean won’s share was even lower, remaining well below both the dollar and partner currencies in terms of the total trade value. That said, won-denominated invoicing shows a gradual increase: the share rose from 1% in 2010 to 2.3% in 2022 for exports, and from 2% to 6% for imports over the same period.

FIGURE 1.

INVOICING CURRENCY SHARE IN EXPORT AND IMPORT: TOTAL TRADE VALUE

Note: Each figure shows the total export value (left) and import value (right) shares by invoicing currency over 2010-2022. DCP refers to dollar-invoiced pricing. In the export figure, LCP denotes pricing in the importing country’s currency, while PCP denotes pricing in Korean won. In the import figure, LCP denotes pricing in Korean won, while PCP denotes pricing in the exporting country’s currency.

The dollar’s dominance is evident not only in the trade value but also in its incidence across trade types. We define a trade type as a destination (or origin) country by product pair at the HSK10 level. As shown in Figure 2, the U.S. dollar is the invoicing currency for the majority of Korea’s export and import trade types. Unlike in the value-based evidence, however, the dollar’s share in export trade types has declined steadily, from 80% in 2010 to about 71% in 2022.

FIGURE 2.

INVOICING CURRENCY SHARE IN EXPORTS AND IMPORTS: TOTAL TRADE TYPES

Note: Each figure shows the total export type (left) and import type (right) shares by invoicing currency over 2010-2022. Trade type is defined as a destination (or origin) country by product pair at the HSK 10 level. DCP refers to dollar-invoiced pricing. In the export figure, LCP denotes pricing in the destination country’s currency, while PCP denotes pricing in Korean won. In the import figure, LCP denotes pricing in Korean won, while PCP denotes pricing in the origin country’s currency.

The decline is largely driven by the rising shares of trade invoiced in partner currencies and in Korean won. For exports, the share invoiced in partner currencies increased from 11% in 2010 to 14% in 2022, while the won-denominated share rose from 8% to 14% over the same period.

We observe a similar pattern for imports. The U.S. dollar’s share fell steadily from 60% in 2010 to about 52% in 2022. Unlike exports, however, the share invoiced in partner currencies fell only slightly, from 32% in 2010 to 30% in 2022. Accordingly, the decline in dollar- and partner-currency invoicing for imports can mainly be explained by the growing share invoiced in Korean won, which increased markedly from 7% in 2010 to 17% in 2022.

This suggests that the won is used primarily in lower-value transactions. As shown in Table 2, the average transaction value per trade type denominated in won for exports is $0.5 million, well below the $2.7 million for U.S. dollar-denominated transactions and the $1.9 million for transactions denominated in trading-partner currencies. For imports, the average transaction value per trade type denominated in won is $1.0 million, similar to the $1.0 million for partner-currency transactions (Table 3). In contrast, the average transaction value for U.S. dollar-denominated imports is $3.8 million, far exceeding the corresponding values for the won and partner currencies. Overall, Korea’s trade patterns indicate that the won is predominantly used for smaller transactions, whereas the U.S. dollar is used for larger ones.

Despite its relatively low average transaction value, the Korean won is also used as an invoicing currency across a wide range of countries (Tables 2 and 3). On average, won-denominated exports involved 190 destination countries, and won-denominated imports involved 172 origin countries – both broadly comparable to the 235 countries covered by U.S. dollar-invoiced transactions. On the other hand, trading-partner currencies were used in transactions with only a limited set of countries.

TABLE 2

OVERVIEW OF EXPORT INVOICING CURRENCIES

Note: All figures were initially computed for each year and then averaged over the 2010–2022 period.

TABLE 3

OVERVIEW OF IMPORT INVOICING CURRENCIES

Note: All figures were initially computed for each year and then averaged over the 2010–2022 period.

Taken together, these patterns indicate that the won is used primarily for lower-value transactions yet is used in a relatively large number of trade types given the amounts involved. These findings offer new evidence pertaining to currency use in a small open economy that lacks vehicle currency status, highlighting how the role of the domestic currency in trade invoicing may be more nuanced than the existing literature suggests. Given that a substantial share of firms continues to invoice in Korean won, conclusions drawn solely from aggregate invoicing data may understate — or at least partially obscure — the relevance of PCP in exports and LCP in imports in a small open economy, particularly for firms that engage in a small volume of transactions.

III. Background

In this section, we examine the determinants of the invoicing-currency choice in Korean exports. We begin by estimating the linear probability model, closely following Amiti et al. (2022).

Here, s denotes products at the HSK 10-digit level, i denotes the product group at HSK 2-digit level, k denotes export destination countries, c denotes the invoicing currency, and t is time defined in year-month. The dependent variable ι_skct is a dummy variable that equals 0 if the invoicing currency is the Korean won and 1 otherwise. captures the share of imported intermediate inputs paid in a foreign currency out of the total intermediate inputs required to produce item s . MKT Share_sk measures the share of destination country k ’s total imports (HS 4-digit level) accounted for by Korea’s export product s . Selections of and MKT Share_sk are motivated by Amiti et al. (2022), who show that the determinants of exchange-rate pass-through (ERPT) play an important role in shaping invoicing-currency choices. In particular, both variables are well-established determinants of the degree of ERPT, as shown in Amiti et al. (2018). It is also known that competitors’ invoicing choices can be important determinants (Mukhin, 2022). Hence, we include ι_−1,sk , which captures the degree of strategic complementarity in invoicing currency decisions, measured as the share of foreign-currency invoicing for other items in the same HS 4-digit level × country market. We construct these control variables and use their averages over the sample period following Amiti et al. (2022). It is motivated by the fact that variations in these controls are driven primarily by cross-sectional differences rather than time-series movements.

Vector X_skt includes additional determinants that may affect the invoicing-currency choice. Specifically, we control for (i) past experience with foreign-currency invoicing, (ii) the size of the foreign-exchange market for currency k relative to the won, (iii) the volatility of the won-currency k exchange rate, and (iv) the average share of foreign-currency debt among firms exporting item s . The inclusion of past experience with foreign-currency invoicing is motivated by Crowley, Han, and Son (2020), who find that firms’ prior invoicing decisions significantly shape their subsequent currency choices. The size of the foreign exchange market and the volatility of the won-currency k exchange rate are included as macro-level determinants of the invoicing-currency choice, following Goldberg and Tille (2016). Finally, based on Kim and Lee (2024) and Kim et al. (2025), we include the foreign currency debt ratio — a factor known to influence the degree of exchange-rate pass-through — as an additional control. We also include month-by-year fixed effects and product-group-by-destination-country fixed effects. Standard errors are two-way clustered by time and export destination country.

We also analyze the determinants of dollar pricing in exports by running the following regression model:

Here, the dependent variable is a dummy variable that takes a value of 1 if the invoicing currency is the U.S. dollar and 0 otherwise. measures the share of imported intermediate inputs invoiced in dollars relative to the total intermediate inputs used to produce item s . MKT Share_sk ptures the share of South Korean exports of item s in the destination market (HS 4-digit level × country). measures the degree of strategic complementarity in the invoicing-currency choice: the dollar invoicing share of other items within the same market (HS 4-digit × country). Vector includes additional controls, including past experience with dollar invoicing, the size of trading partner k ’s foreign-exchange market relative to the dollar, and the volatility of the dollar–trading-partner currency exchange rate. The volatility of the won/trading partner currency exchange rate and the average proportion of foreign currency debt among firms exporting a specific product s are also included. Control variables are constructed in the same way as in the won-invoicing choice analysis. We include month-by-year fixed effects and product-group-by-destination-country fixed effects. We report two-way clustered standard errors, clustered by time and export destination country.

The share of foreign currency intermediate goods to produce item s and that of dollar intermediate goods are constructed using the IO-Table provided by the Bank of Korea and Customs data. We use concordance tables that map I–O product categories to HSK 6-digit codes to map the I–O items to the customs (HSK code) classification. MKT Share_sk is computed using UN-Comtrade data on total imports by country and product. ι_−1,sk and are constructed as value shares. For a given market, they equal the share of the transaction value of other products (excluding the s product) invoiced in a foreign currency and in U.S. dollars among Korean exporters, respectively. Past experience with foreign-currency (or dollar) invoicing is measured as the share of exports of item s that were invoiced in a foreign currency (or in U.S. dollars), up to and including the previous year. The foreign-exchange market size of trading partner k ’s currency relative to the U.S. dollar and relative to the Korean won is constructed using BIS data, specifically the Triennial Central Bank Survey of Foreign Exchange and OTC Derivatives Markets. Because the survey is conducted every three years, we convert the series to an annual frequency by means of linear interpolation.

We measure the volatility of USD/trading partner and KRW/trading partner exchange rates using the coefficient of variation of monthly exchange-rate changes over a five-year window centered on the reference month – that is, the 30 months preceding and the 30 months following it. Exchange-rate data are obtained from the IMF and the BIS.

We compute the average share of foreign-currency debt using KIS-Value data. Using a concordance table that maps each item (HSK 4-digit) to firms’ industries (KSIC Rev. 10), we construct, for each item, the sales-weighted average short-term foreign-currency debt ratio among firms producing that item s .

Table 4 reports the baseline results for won invoicing choices. First, a higher share of foreign currency imported intermediate inputs is associated with a lower probability that the corresponding item is invoiced in won. Specifically, a one percentage point increase in increases the probability that export prices are denominated in foreign currency by 0.08 percentage points, and this effect is statistically significant. Given that the average value of is 35% (standard deviation: 16%), the estimated coefficient implies that there exists a sizable impact on the likelihood of won invoicing. This result is consistent with theoretical predictions that the use of foreign-currency intermediate inputs – an important determinant of exchange-rate pass-through – shapes exporters’ invoicing choices. It can also be interpreted as a form of natural hedging against exchange-rate-induced fluctuations in marginal production costs. The average share of foreign-currency debt among firms exporting each item s is also positively related to foreign-currency invoicing. Specifically, a one percentage point increase in the share of foreign-currency debt raises the probability of pricing in a foreign currency by 0.07 percentage points, and the estimate is statistically significant at the 5% level. The coefficient of foreign-currency debt is slightly smaller than that of foreign-currency intermediate inputs. Given that the average foreign-currency debt share is 12% (standard deviation: about 8%), its overall effect is likely more modest than that of imported intermediate-input exposure. Firms’ foreign-currency borrowing decisions may reflect a range of firm-specific characteristics, raising concerns about potential endogeneity. However, in our data, foreign-currency liabilities exclude accounts payable, making it unlikely that this measure mechanically captures borrowing linked to imported intermediate inputs. In the baseline specification, we also control for , meaning that endogeneity concerns arising specifically from imported intermediate-goods exposure should be limited. In addition, we use either the period average of foreign-currency debt or a lagged measure as a robustness check, which helps mitigate concerns about reverse causality. That said, omitted-variable bias may remain if an unobserved factor jointly influences both the invoicing currency and foreign-currency borrowing. Nevertheless, the result is informative in that it provides empirical evidence that firms’ financial positions and trade-related currency choices are closely related.

TABLE 4

DETERMINANTS OF WON INVOICING

Note: The dependent variable ι_{𝑠𝑘𝑐𝑡} is a dummy variable that equals 0 if the invoicing currency is the Korean won and 1 otherwise. We include month-by-year fixed effects and product-group-by-destination-country fixed effects. Two-way clustered standard errors, clustered by time and export destination country, are reported in parentheses. *p<0.1, **p<0.05, ***p<0.01.

The likelihood of setting a foreign currency as the invoicing currency depends heavily on past experience. Specifically, a one percentage point increase in the share of past foreign-currency invoicing for a given item is associated with a 0.27 percentage point increase in the probability that the item is currently invoiced in a foreign currency. This estimate is statistically significant at the 1% level. This result is consistent with earlier findings (Korea International Trade Association, 2014) that firms primarily rely on past practices when setting export prices. Additionally, strategic complementarity is also confirmed to be an important factor influencing the invoicing currency decision. A one percentage point increase in the proportion of competitors invoicing in a foreign currency raises the probability of invoicing in a foreign currency by 0.52 percentage points, and this estimate is also statistically significant. This indicates that firms tend to align their export invoicing currency with competitors’ currencies when setting their prices, consistent with strategic complementarity in invoicing-currency choices as found in prior research (Amiti et al., 2022; Mukhin, 2022).

We do not find evidence that the share of Korean exports in the destination country’s total imports MKT Share_sk , the size of the trading partner’s foreign exchange market relative to the won, or the volatility of the won-trading partner exchange rate is associated with the choice of invoicing currency.

Table 5 reports the results from the dollar-invoicing (dollar-pricing) determinants model. Similar to the won-invoicing determinants, a higher share of dollar-denominated imported intermediate inputs is associated with a stronger tendency to invoice in dollars. Specifically, a one percentage point increase in the share of dollar-denominated intermediate inputs raises the probability of dollar invoicing by 0.17 percentage points, and this effect is statistically significant. The likelihood of invoicing exports in U.S. dollars is also found to depend strongly on past experience. Specifically, a one percentage point increase in an item’s historical dollar-invoicing share is associated with a 0.16 percentage point increase in the probability that the item is currently invoiced in dollars, with this estimate also being statistically significant at the 1% level. Similar to won invoicing, dollar invoicing is also significantly influenced by competitors’ currency choices. In export markets defined at the HS 4-digit by country level, a one percentage point increase in the share of dollar-denominated trade in other products (excluding product s ) is associated with an approximately 0.46 percentage point increase in the probability that product s is invoiced in U.S. dollars.

TABLE 5

DETERMINANTS OF DOLLAR INVOICING

Note: The dependent variable is a dummy variable that takes a value of 1 if the invoicing currency is the U.S. dollar and 0 otherwise. We include month-by-year fixed effects and product-group-by-destination-country fixed effects. Two-way clustered standard errors, clustered by time and export destination country, are reported in parentheses. *p<0.1, **p<0.05, ***p<0.01.

The determinants of dollar invoicing differ in several respects from those of won invoicing. First, the effect of MKT Share_sk is estimated to be negative and statistically significant. The results indicate that a higher market share in a given destination market is associated with a lower propensity to invoice in U.S. dollars. This pattern is consistent with the findings of Amiti et al. (2022): a higher market share reflects greater markup elasticity, causing firms to optimally choose local-currency pricing to limit price volatility in the destination market. Second, unlike won invoicing, dollar invoicing is influenced by the depth of the foreign exchange market. Specifically, the larger the destination currency’s foreign exchange market relative to that of the U.S. dollar, the lower the likelihood of dollar invoicing. Quantitatively, a one percentage point increase in the relative size of the destination currency’s foreign exchange market reduces the probability of dollar invoicing by 0.9 percentage points, and this effect is statistically significant at the 5% level. Deeper foreign exchange markets typically imply lower transaction costs and greater liquidity, making invoicing in non-dollar currencies more attractive. Exchange rate volatility is also a significant determinant. Greater volatility in the won–partner currency exchange rate is associated with a higher likelihood of using a third invoicing currency – specifically the U.S. dollar. In contrast, greater volatility in the dollar-partner currency exchange rate is associated with a lower likelihood of dollar invoicing. Goldberg and Tille (2016) argue that such dollar share shifts are driven by adjustments in PCP; however, as indicated in Table 4, we find no significant adjustment in won invoicing in response to these factors. These findings suggest that the crowding-out of PCP emphasized in the existing literature may be less applicable to Korea, a small open economy whose currency does not serve as a vehicle currency. They also imply, albeit cautiously, that efforts to expand the foreign exchange market size or stabilize exchange rates may not by themselves lead to a substantial increase in won invoicing.

Finally, the share of foreign-currency (FC) debt — an important determinant of won invoicing — does not appear to be statistically significant for dollar invoicing. This may reflect heterogeneous effects of FC debt across firms. For instance, Kim and Lee (2024) and Kim et al. (2025) suggest that the impact of FC debt on exchange-rate pass-through differs by firm size: higher FC debt is associated with more incomplete pass-through among smaller exporters but is associated with more complete pass-through among larger exporters. Because PCP transactions tend to involve relatively small firms (Table 2), a higher share of FC debt in this group may be associated with a greater likelihood of dollar or LCP invoicing rather than PCP, which is broadly consistent with the positive coefficient of non-PCP in Table 4.

On the other hand, for DCP — where larger firms are more prevalent — these offsetting effects may make the estimates less precisely identified.

IV. Differential Effects of Exchange Rates by Invoicing Currency

This chapter examines the macroeconomic implications of invoicing currency. In the short run, when nominal prices are relatively sticky, the impact of the exchange rate on exports depends heavily on the invoicing currency. Over the medium to long run, however, the role of the invoicing currency diminishes as firms have enough time to adjust prices in response to exchange rate changes. Accordingly, this chapter investigates how exchange-rate pass-through – conditional on the invoicing currency – evolves over time. First, we assess the effects of exchange rates on export prices, and then examine the corresponding response of export quantities.

A. Export Price

1. Empirical model and data

This section analyzes three main invoicing currencies – the Korean won (PCP), the U.S. dollar (DCP), and the trading partner’s currency (LCP) – following the framework of Amiti et al. (2022). For firm f that cannot adjust prices flexibly, the realized price expressed in the trading partner’s currency (*) can be expressed as follows:

Here, denote prices set in advance, before economic conditions are realized. Accordingly, the change in the realized price expressed in the trading partner’s currency can be written as follows:

In this equation, denotes the log change in price when firm f can flexibly choose the optimal price given current conditions, while represents the log change in the exchange rate between trading-partner currencies k and l . The parameter δ indicates the probability that prices cannot be adjusted. The expected value of the realized price denominated in the trading partner’s currency can be rewritten as follows:

where ι^L and ι^D are dummy variables for LCP and DCP, respectively. The PCP indicator can be defined as ι≡ι^L+ι^D . Consistent with the previous chapter, ι can therefore be interpreted as a dummy variable that equals 0 when PCP is chosen.

Based on the theoretical predictions (5), we specify the following empirical model to estimate how exchange-rate movements affect export prices differently across invoicing currencies.

Here, p_skct denotes the log price of export item s , expressed in the trading partner k ’s currency. Item s is defined at the 10-digit HS level; c indicates the invoicing currency; and t denotes the year–month. ι_skct and are dummy variables for PCP and DCP. e_kt is the nominal exchange rate between Korea and trading partner k , expressed as units of trading partner k ’s currency per the Korean won (i.e., the value of the won). is the nominal exchange rate between the United States and trading partner k , expressed as units of trading partner k ’s currency per the U.S. dollar. Δ_hP_skct denotes the log difference between prices at t−h and t . Δ_he_kt Δ_h are defined analogously. We consider horizons h ranging from two to 24 months. For each horizon h , we estimate the regression separately to capture the dynamic, invoicing-currency-specific exchange-rate effects on export prices.

Accordingly, the key parameters of interest are the horizon-specific coefficients γ_h and for each h . These coefficients can be interpreted as the degree of price rigidity at horizon h , while also reflecting how exchange-rate pass-through differs by invoicing currency. The theoretical model predicts the relationship between γ_h and as −γ_h≈ > 0 , also predicting that the magnitude of each coefficient declines over time as prices adjust.

The degree of exchange-rate pass-through may also be shaped by factors beyond the invoicing currency. Moreover, as noted by Amiti et al. (2022), pass-through determinants may also influence firms’ invoicing-currency choices. To mitigate endogeneity concerns, we control for a rich set of fixed effects (time by country by product) that absorb factors affecting firms’ optimal price adjustments in response to exchange-rate movements. Given potential dynamic dependence in the regression, we additionally control for the lagged dependent variable Δ_hP_skct−1 . Finally, we use two-way clustered standard errors, clustering by time and by destination country.

The sample covers January of 2010 through December of 2019, i.e., the pre COVID-19 period. The dataset reports export values (in U.S. dollars) and export weights (in kilograms) by item but does not provide prices directly. Following the literature, we therefore use the unit value – defined as export value divided by export weight – as a proxy for export prices. Because items are recorded at the 10-digit HSK level, unit values should provide a reasonably informative measure of export prices. Nonetheless, unit values may still be biased if the product composition or quality varies within an HSK 10-digit category. Accordingly, consistent with prior work (Amiti et al., 2022), we exclude observations for which the annual change in unit value exceeds 3 or falls below 1/3.

2. Empirical results

Figure 3 illustrates how the estimated coefficient γ_h evolves over time. Consistent with the theoretical prediction, the coefficients are estimated as negative. This implies that when exports are invoiced in the trading partner’s currency (LCP) rather than in the Korean won (PCP), exchange-rate pass-through into prices denominated in the trading partner’s currency is lower. At the 12-month horizon, the estimated pass-through differs by approximately 44 percentage points between won-invoiced and non-won-invoiced transactions. The results also show that the difference declines over time. While the estimated gap is roughly 63 percentage points at the two-month horizon, it declines steadily (in absolute value) at longer horizons, falling to roughly 33 percentage points at 24 months. Taken together, these findings suggest that the invoicing currency plays a central role in shaping export-price responses in the short run – when nominal prices are relatively sticky – but becomes less important over the medium run as firms gradually adjust their prices to exchange-rate changes.

FIGURE 3.

DIFFERENTIAL EXCHANGE-RATE EFFECTS ACROSS INVOICING CURRENCIES: PCP VS OTHERS

Note: The solid red line represents the coefficient estimate γ_h in Equation (6), which captures how the differential effect of exchange rate movements on won-invoiced transactions evolves over time (by month). The negative estimate reflects higher exchange-rate pass-through for won-invoiced transactions in response to changes in the won-destination currency exchange rate. The dashed lines denote the 90% confidence interval based on two-way clustered standard errors, clustering by time and by destination country.

Figure 4 plots the estimated coefficient , which captures the differential effect of dollar invoicing on exchange-rate pass-through. The coefficient is estimated to be positive, indicating that when exports are invoiced in U.S. dollars, pass-through from the dollar-trading partner currency exchange rate to prices is significantly higher. At the 12-month horizon, the estimated pass-through is about 47 percentage points higher for dollar-invoiced transactions than for non-dollar-invoiced transactions. This differential narrows over time: it is roughly 74 percentage points at the two-month horizon and declines to about 34 percentage points at the 24-month horizon.

FIGURE 4.

DIFFERENTIAL EXCHANGE-RATE EFFECTS ACROSS INVOICING CURRENCIES: DCP VS OTHERS

Note: The solid red line represents the coefficient estimate in Equation (6), which captures how the differential effect of exchange rate movements on dollar-invoiced transactions evolves over time (by month). The positive estimate reflects higher exchange-rate pass-through for dollar-invoiced transactions in response to changes in the dollar-destination currency exchange rate. The dashed lines denote the 90% confidence interval based on two-way clustered standard errors, clustering by time and by destination country.

The coefficients γ_h and , which capture the differential exchange-rate effects by invoicing currency, are closely related and primarily reflect price rigidity. While the degree of price rigidity may vary across invoicing currencies, the model implies that the two coefficients should be of similar magnitude. Accordingly, we use the model’s implications to formulate the following hypothesis and test this prediction empirically.

Figure 5 reports the test of the hypothesis described above. The solid red line plots the estimated value of γ_h + , and the gray dotted lines indicate the 90% confidence interval. γ_h and are similar in absolute magnitude: at most horizons h , we cannot reject the null that their sum equals zero. This finding is consistent with the model’s prediction that both coefficients primarily reflect price rigidity and should therefore be comparable in magnitude.

FIGURE 5.

SUM OF ESTIMATED PARAMETERS: 𝛾 + 𝛾^𝐷

Note: The solid red line plots the estimated value of 𝛾_ℎ + in Equation (7), while the dashed lines denote the 90% confidence interval based on two-way clustered standard errors, clustering by time and by destination country.

B. Export Quantity

1. Empirical Model

The degree of exchange-rate pass-through is found to vary with the invoicing currency. This heterogeneity matters because exchange rates influence exports’ relative prices. When the relative prices of otherwise similar goods differ simply due to the invoicing currency, this can affect resource allocation within the economy. Hence, the transmission of external shocks – and the appropriate policy response – may depend on the currency in which export prices are invoiced.

For a more complete assessment of the macroeconomic implications of the invoicing-currency choice, we therefore examine whether and how these price effects translate into changes in real outcomes.

The baseline regression model follows Amiti et al. (2022), as follows:

The regression (8) is the second-stage estimating equation. The first stage corresponds to the specification (6) in the previous section, which estimates invoicing-currency-specific exchange-rate pass-through into export prices. Accordingly, serves as an instrument for export prices: it is the predicted component of price changes implied by exchange-rate movements and the invoicing currency.

The dependent variable y_skct is the log export quantity, and Δ_hy_skct denotes the change in the log export volume from t−h to t . s indexes products at the 10-digit HSK level, c denotes the invoicing currency, and t is the year-month. Because export quantities may be influenced by sectoral characteristics and macroeconomic conditions, we include time × destination-country × product (HSK 10-digit) fixed effects, as in the previous section. With these controls, our estimates can be interpreted as capturing how price changes affect export quantities, holding constant common equilibrium conditions for a given product and destination market.

2. Empirical Results

Figure 6 shows that the invoicing currency plays an allocative role. The estimated coefficients are negative across all horizons, consistent with the standard inverse relationship between prices and demand.

FIGURE 6.

ELASTICITY OF EXPORT QUANTITIES TO EXPORT PRICES

Note: The solid red line represents the coefficient estimate 𝛽_ℎ in Equation (8), which captures how the invoicing-currency-specific exchange-rate effects translate into changes in export quantities over time (month). The negative estimate reflects a standard downward slope demand curve. The dashed lines denote the 90% confidence interval based on two-way clustered standard errors, clustering by time and by destination country.

The estimated price elasticity of export quantities is smaller at short horizons but increases gradually in absolute value over time. Specifically, it is -0.30 at the two-month horizon, -0.67 at 12 months, and -0.83 at 24 months. This pattern suggests that export quantities adjust only gradually to price changes, indicating rigidity not only in prices but also in quantity responses. Such sluggish adjustments may reflect the long-term trading relationships, which can dampen short-run quantity reallocation. This pattern is consistent with a J-curve dynamic, in which prices react relatively quickly to exchange-rate changes while quantities adjust with a lag.

The medium- to long-run elasticity estimate (-0.83) is smaller in absolute value compared to the corresponding values reported in Amiti et al. (2022) and in other studies using aggregate national data (often around -2.0). This result may be driven by the structure of Korea’s exports, which are concentrated in sectors with low price responsiveness. Table 6 indicates that Korea’s key export products tend to exhibit relatively low price elasticities.

TABLE 6

ELASTICITY OF EXPORT QUANTITIES BY BROAD PRODUCT CATEGORIES

Note: Each coefficient estimate represents product-level export quantity elasticity over a 24-month horizon. We compute two-way clustered standard errors, clustering by time and by destination country. *p<0.1, **p<0.05, ***p<0.01.

Elasticities were computed at the product level by initially grouping the full sample into MTI categories and then estimating the two-stage model separately for each category. The estimated elasticities are negative for all products, and all are statistically significant at the 5% level except for one product - flat panel displays. These results confirm substantial heterogeneity in price elasticity across export items. Notably, the export-volume elasticity for semiconductors – Korea’s key export product – is estimated to be -0.7, smaller in absolute value than for many other products. Wireless communication devices and flat panel displays appear particularly inelastic. For several items, the estimated elasticity is close to one in absolute value, but the results overall indicate that many of Korea’s major export products exhibit relatively low levels of price elasticity.

C. Additional Exercises

The product-level export-quantity elasticities in Table 6 provide additional insight into the determinants of the invoicing-currency choice. When demand is more elastic, firms have stronger incentives to limit price changes due to exchange rate fluctuations, which may increase the importance of competitors’ invoicing choices. For example, suppose a firm invoices in the trading partner’s currency while its competitors invoice in U.S. dollars. If the trading partner’s currency appreciates against the dollar, the firm’s price – relative to dollar-invoiced competitors – would rise in the destination market, potentially reducing its competitiveness. In this case, if demand is highly elastic relative to price, the firm could face a sizable decline in demand – and thus profits. To avoid such outcomes, firms may have an incentive to invoice in U.S. dollars, and this incentive should be stronger for products with greater price elasticity. Consistent with this intuition, the analysis below shows that product-level demand elasticity is systematically related to the invoicing-currency choice.

Table 7 shows the results after augmenting the baseline regression Equation (1) to include (i) product-specific elasticities and (ii) the interactions of product-specific price elasticities with competitors’ invoicing currencies. Given that a sufficiently large sample is required to estimate elasticities, we construct elasticity measures at the HS 2-digit level. This analysis therefore includes only time and destination fixed effects.

As shown in the first two columns of Table 7, the interaction term between elasticity and the foreign-currency invoicing share of other items is estimated to be positive. The result indicates that for more elastic products, invoicing choices co-move more closely with the foreign-currency invoicing share of competing products. Put differently, when demand is more elastic, firms have stronger incentives to limit exchange-rate-driven price fluctuations, making competitors’ invoicing practices more influential.

The remaining two columns of Table 7 report analogous results for U.S. dollar invoicing. Consistent with previous results, the interaction term is estimated to be positive, implying that the effects of the dollar invoicing share of other competing products on a product’s own dollar invoicing choice increase with elasticity.

TABLE 7

ADDITIONAL DETERMINANTS OF INVOICING DECISIONS: ELASTICITY

Note: The dependent variable 𝜄_{𝑠𝑘𝑐𝑡} in columns (1) and (2) is a dummy variable that takes a value of 0 if the invoicing currency is the Korean won and 1 otherwise, while the dependent variable in columns (3) and (4) is a dummy variable that takes a value of 1 if the invoicing currency is the U.S. dollar and 0 otherwise. In columns (1) and (2), we additionally control for 𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦_𝑠 and 𝜄_{−1,𝑠𝑘}×𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦_𝑠 in regression Equation (1). In columns (3) and (4), we additionally control for 𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦_𝑠 and ×𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦_𝑠 in regression Equation (2). We include month-by-year fixed effects and product-group-by-destination-country fixed effects. Two-way clustered standard errors, clustered by time and export destination country, are reported in parentheses. *p<0.1, **p<0.05, ***p<0.01.

D. Aggregate Implications for the Korean Economy

Next, we examine how exchange rate movements shape aggregate outcomes in the Korean economy, focusing on episodes of pronounced global dollar strength — those surrounding Russia’s invasion of Ukraine in 2022 and the subsequent monetary tightening in the United States. As shown in Figure 7, the U.S. dollar appreciated sharply on a year-on-year basis during this period, reaching a peak in October of 2022.

FIGURE 7.

EXCHANGE RATE DYNAMICS DURING THE RUSSIAN-UKRAINE WAR

Note: The solid blue line represents the Korean Nominal Effective Exchange Rate (NEER) and the solid orange line represents the Dollar Index. Both series are normalized to 100 in January of 2020.

In standard open-economy frameworks, depreciation of the Korean won is expected to improve price competitiveness and stimulate export volumes. However, this prediction appears at odds with the data: in October of 2022, Korea’s export volume declined by 3.2% year-on-year despite a substantial depreciation of the won. While export dynamics are influenced by a range of macroeconomic factors, this pattern suggests that the global strong-dollar episode itself may have dampened export performance.

To quantify this mechanism, we construct two counterfactual scenarios that isolate the role of the invoicing currency. In the first scenario, all exports currently invoiced in U.S. dollars (DCP) are instead assumed to be invoiced in the importer’s currency (LCP). In the second, all exports are assumed to be invoiced in Korean won (PCP). The baseline corresponds to realized outcomes.

The results indicate that the invoicing currency plays a quantitatively meaningful role. As shown in Figure 8, under the LCP counterfactual, the year-on-year growth rate of export volumes in October of 2022 would have been 3.05 percentage points higher than in the baseline. Under the PCP counterfactual, the effect is even larger: export growth would have been higher by 5.24 percentage points in October of 2022. Consistent with the findings in Figures 3 and 4, the effects weaken over time.2

FIGURE 8.

COUNTERFACTUAL EXPORT GROWTH RATE

Note: The blue and red bars represent the export volume growth of the LCP and PCP scenarios, respectively, relative to the baseline.

Taken together, these findings highlight a key implication of currency invoicing for the Korean economy. When a large share of trade is invoiced in U.S. dollars, broad-based appreciation of the dollar can suppress export volumes — even in the presence of any depreciation of the Korean won. More broadly, the results underscore how the macroeconomic effects of exchange rate movements depend critically on their underlying source. In particular, a case of depreciation driven by global dollar strength may be considerably less expansionary — and may even be contractionary — compared to a case of depreciation driven by idiosyncratic weakness of the Korean won.

V. Conclusion

This paper studies invoicing-currency choices in Korean exports by identifying the key determinants of such choices and analyzing the corresponding implications regarding exchange-rate pass-through and aggregate export dynamics. The results show that the determinants of exchange-rate pass-through – intermediated imported inputs, market share, and foreign currency liability – play an important role in shaping currency invoicing decisions. Invoicing choices also exhibit strong persistence over time and strategic complementarity, with currency choices moving closely with competitors’ invoicing practices within narrowly defined markets. Demand elasticity plays a significant role in amplifying strategic complementarity. With regard to dollar invoicing, certain destination-market characteristics matter as well; the depth and liquidity of destination currency markets and exchange-rate volatility are found to be important determinants of invoicing-currency choices.

These invoicing patterns have significant consequences for exchange-rate transmissions. Export-price pass-through differs sharply across invoicing currencies at short horizons — consistent with sticky prices — but these differences diminish over time as prices adjust. Export quantities also respond gradually to exchange-rate-driven price changes, implying that the choice of invoicing currency has real consequences concerning trade adjustments.

Our counterfactual analysis quantifies the role of the invoicing structure: had all dollar-invoiced exports instead been invoiced in the importer’s currency, export volume growth in October of 2022 would have been 3.05 percentage points higher. Under universal won invoicing, the gain would have been even larger, at 5.24 percentage points. These findings suggest that the prevailing dollar-invoicing structure meaningfully dampened Korea’s export response during the aforementioned strong-dollar episode.

More broadly, the results underscore the importance of distinguishing between sources of exchange rate variation when assessing external vulnerabilities. In a world of dominant dollar pricing, the relevant metric for evaluating export competitiveness is not simply the bilateral won-dollar rate or the nominal effective exchange rate but is the invoicing-currency-specific exchange rates faced by Korean exporters across destination markets. Monitoring and modeling this finer structure of exchange rate exposure should be a priority for policymakers.

Overall, this paper provides a comprehensive empirical analysis of how exchange-rate movements transmit to a small open economy operating under dominant currency pricing, using transaction-level micro data from Korea. The findings speak to a broader set of economies that do not issue vehicle currencies yet face pervasive dollar invoicing in their trade. As global monetary conditions continue to generate sharp swings in the U.S. dollar, the invoicing-currency dimension of exchange rate transmission deserves greater attention in the areas of both academic research and policy design.