BIS Working Papers
Industry heterogeneity and exchange rate pass-through
by Camila Casas
Monetary and Economic Department
JEL classification: D22, D84, E31
Keywords: inflation expectations, firms' survey, new information.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2019. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISSN 1682-7678 (online)
Industry heterogeneity and exchange rate pass-through
by Camila Casas
In the presence of price rigidities, nominal exchange rate fluctuations can have real effects on the economy. External shocks may have differentiated effects across economic sectors depending on firms’ marginal cost structure and features of the demand they face, such as strategic complementarities. I analyze the relationship between the exchange rate pass-through into export and import prices and volumes and the use of imported inputs in production, an important determinant of marginal cost. Using microdata from Colombia, I show that manufacturing industries differ significantly in their use of imported inputs and in the estimated exchange rate pass-through. I find a clear correlation between the use of imported inputs and the response of prices to changes in exchange rates. That is, the exchange rate pass-through into prices tends to be larger for industries in which firms use a larger share of imported inputs. The link is stronger in the case of exports, but the effect on the pass-through into import prices is also positive. In contrast, I do not find a clear correlation between the use of imported inputs and the response of traded quantities to changes in exchange rates.
JEL classification: D22, D84, E31
Keywords: inflation expectations, firms' survey, new information.
In the presence of price rigidities, fluctuations in nominal exchange rates can have real effects on a country’s economy. Changes in a country’s currency price can impact output, prices and trade flows, and the effects will vary according to the nature of the price rigidities. Furthermore, external shocks may have differential effects across economic activities depending on firms’ productive structures and the particular features of the demand they face. Two of the structural determinants of the exchange rate pass-through into export and import prices are the use of imported inputs in production and the ability of firms to charge variable markups across markets and over time (Goldberg and Hellerstein (2008)). Given that variations in the effects of international shocks across economic agents can create winners and losers and a reallocation of productive factors may take place, it is important for policymakers to understand any potential differences and their underlying causes.
In this paper I analyze the response of export and import prices and volumes to exchange rate changes. First, I use detailed firm-level operational data to examine two potential sources of heterogeneity across 2-digit manufacturing industries in the use of imported inputs: differences in the share of imported inputs in firms’ total input purchases, and differences in the elasticity of output with respect to intermediate inputs that reflects the relative importance of inputs for a firm’s productive process. Then, I use transaction-level trade data to estimate the pass-through of exchange rate fluctuations into export and import prices for each industry, and I analyze the relationship between these estimates and the use of imported inputs at the sector level.
Finally, I study the effects of exchange rate changes on traded quantities to understand the extent of expenditure switching in different sectors.
My main results can be summarized as follows. First, there are significant variations in industries' use of imported inputs that can be explained mainly by differences in the share of imported inputs in firms' total input purchases. While domestic inputs account for over 90% of total purchases in some industries, other sectors import over 30% of their inputs. Second, the exchange rate pass-through into export prices varies significantly across industries; variations in the exchange rate pass-through into import prices are smaller. Third, the pass-through estimates correlate positively with the use of imported inputs at the 2-digit manufacturing industry level: industries with a larger share of imported inputs tend to have a higher exchange rate pass-through into both export and import prices. The correlation appears, however, to be stronger for export prices, since fluctuations in the exchange rate have a direct effect on the marginal cost of production of firms using imported inputs. In the case of imports, the exchange rate has an indirect effect on prices through local competitors' marginal cost due to the presence of strategic complementarities. Finally, the response of export quantities to changes in the exchange rate varies widely across industries, whereas the effect on imports is fairly uniform. Changes in exported volumes are found to be almost negligible for some industries, while a peso depreciation is associated with a decline in exports across destinations for others. In the case of imports, albeit heterogeneous in magnitude, the estimated effect of a peso depreciation is negative for the manufacturing sector as a whole and for almost all individual industries. I do not find a clear correlation between the estimated elasticities of traded quantities and the use of imported inputs.
This paper aims to contribute to the extensive literature on incomplete pass-through and its determinants. Several studies have shown that exchange rate fluctuations are only partially transmitted to the prices of traded goods. For instance, Campa and Goldberg (2005) show that pass-through into import prices for the OECD countries is, on average, close to one-half. Using a larger sample of countries that also includes emerging economies, Bussiere, Delle Chiaie and Peltonen (2014) show that passthrough is incomplete for both export and import prices, and that the two are correlated. And for Colombia, Rincon-Castro and Rodriguez-Nino (2018) show that pass-through into import prices is incomplete along the distribution chain. These papers emphasize the role of macroeconomic factors such as inflation, the stability of the exchange rate and trade composition.
A strand of the literature has used microdata to focus on the link between pass-through and microeconomic variables. Among firm-level variables that have been found to affect pass-through are the currency of invoicing (Gopinath, Itskhoki and Rigobon (2010)); market structure and the ability of exporters to charge variable markups (Amiti, Itskhoki and Konings (2018); Atkeson and Burstein (2008); Auer and Schoenle (2016); Fitzgerald and Haller (2014)); and features that affect marginal cost. These include the use of imported inputs in production (Amiti, Itskhoki and Konings (2014); Fauceglia, Shingal, and Wermelinger (2014)) heterogeneous productivity (Berman, Martin and
Mayer (2012); Garetto (2016)); heterogeneous levels of quality (Chen and Juvenal (2016)); and distribution costs (Burstein, Neves and Rebelo (2003)).
The widespread use of global value chains has put a spotlight on the role of imported inputs in determining trade prices and volumes. The paper contributes to this topic in several ways. First, my results for import prices complement those for export prices in Amiti, Itskhoki and Konings (2014) and Fauceglia, Shingal and Wermelinger (2014). As Amiti, Itskhoki and Konings (2018) show, firms adjust their prices in reaction to their competitors’ price changes. In this context, exchange rate shocks affecting the marginal cost of local producers using imported inputs can affect not the price of their exports, but also the price of imports competing with them. I explore this indirect channel on import prices. Second, although I use detailed firm-level data for my estimations, the relationship between the use of imported inputs and the exchange rate pass-through is examined at the industry level. The sector-level analysis can be useful for policymakers whenever detailed firm-level data is not available. A third contribution is the analysis of two separate sources of heterogeneity in the effects of imported inputs on trade prices and volumes. The literature has studied the share of imports in total input purchases as a source of heterogeneity in firms’ marginal cost. Besides that, I analyze the heterogeneity generated by differences across sectors in the elasticity of output with respect to intermediate inputs; that is, in the relative importance of inputs for firms' production.
The remainder of the paper is organized as follows. Section 2 presents the theoretical background that guide the empirical strategy. Section 3 describes the data sources for the empirical analysis. Section 4 contains sectoral statistics on the share and use of imported inputs in production, regression results on exchange rate pass-through, and the link between these two sets of results. Finally, Section 5 concludes.
The contribution of this paper is essentially empirical, but the work is guided by the Dominant Currency Paradigm (DCP) model developed in Casas, Diez, Gopinath and Gourinchas (2017). Based on the evidence from granular trade data, the model incorporates three key features. First, regardless of the origin or destination of trade, firms set their prices in a dominant currency and change them infrequently. Second, firms use both domestic and imported intermediate inputs for production. And third, firms face strategic complementarities and set flexible markups that vary across time and markets (see Casas et al. (2017) for the details). In this section I present the key elements of the model to show how the key variables relate to each other and thus set the course for the empirical strategy.
1.1. The Model
Consider a three-country world: a small open economy (Я, home) and two large countries (U, the dominant currency country, and R, the rest). H trades goods and assets with U and R, and from the perspective of H the traded quantities and prices of U and R are exogenous.
Consumers have a per-period utility function that depends on their consumption of a bundle of goods and on their labor supply,
where о is the coefficient of relative risk aversion; ф is the inverse of the Frisch elasticity of labor supply; and к scales the disutility of labor. C is assumed to be a Kimball homothetic demand aggregator implicitly given by
where CiH is the consumption in H of variety w produced in country i e [H, U, R}; П is a measure of the varieties produced in i; and yi is a parameter that measures the home bias in H, with = 1. These preferences are a way of introducing strategic complementarities (and variable markups) since, with the resulting kinked demand functions, producers do not want to deviate from their competitors’ prices.
The optimality conditions of the household's problem yield the following demand system:
aggregate demand, and PiH, t(w) denotes the home price of variety ш produced in country i and sold in H.
Some further notation is useful for the empirical predictions. Define the elasticity of demand
Markups and their elasticities are important factors in the empirical analysis presented below, since they affect how production costs changes that result from exchange rate movements are transmitted to firm-level export and import prices.
Each H producer manufactures a unique variety , combining intermediate inputs and labor according to a Cobb-Douglas production function,
where a is the constant share of intermediates in production and at is an idiosyncratic productivity shock. The intermediate input aggregator Xt takes the same form as the consumption aggregator in equation (2), and the labor input Lt is a CES aggregator of the individual varieties supplied by households.
Each firm’s variety ш is sold both domestically and internationally and its output is used both for final consumption and as an intermediate input for production. Prices for the three markets, H, U and R, are chosen by the firms in a Calvo pricing environment: they can be reset infrequently with a probability (1 — Sp) in any given period. Exported goods can be priced in the dominant currency, or that of the producer, or in the local/destination currency, while domestic prices and wages wt are set in H currency. Let t(ы) denote the price of variety ш produced in H, sold in country i, invoiced in currency j.
Assuming that markets are segmented, the per-period profits of the domestic firm producing variety ш are then given by
where £j t is the nominal exchange rate (expressed as Home currency per unit of foreign currency j); t (w) = t (w) + XJHi t(w) is the demand for ш in country i invoiced in currency j, both used for consumption and as an input in production; Yt(w) = Yii,j t (ы) is the total demand across markets; and MC is the marginal cost given (in logs) by mcJHt = (1 — a)wt + a^iYi (Prn.t + ej,t)- As is usual in the literature, lowercase letters denote the log of the corresponding variables.
Closing the Model
In order to evaluate the transmission of exchange rate fluctuations to prices, the model requires additional conditions on the process for setting interest rates. H’s monetary authority is assumed to set the domestic risk-free interest rate according to an inflation targeting Taylor rule with inertia, and the dominant currency interest rate at which H borrows internationally is assumed to be an increasing function of its external debt.
Given market clearing conditions for goods, labor and bond markets, the model is solved by log-linearizing around a symmetric zero-inflation steady state.
1.2. Exchange Rate Pass-Through
The DCP model delivers a number of novel predictions regarding the effects of exchange rate fluctuations on export and import prices, that is, the exchange rate pass-through (ERPT). The predictions presented in this subsection will be tested in the empirical analysis.
In the general case, firms can adjust prices with a probability of (1 — Sp). The optimal price-reset condition for domestic firms selling in country i and invoicing in currency j implies that export and import price inflation for that specific combination of origin, destination, and currency of invoicing is given by
respectively (see the appendix in Casas et al. (2017) for the derivation). Here, Ap = (1 — Sp)(1 — pSp)/Sp; mcJit is the log of the nominal marginal cost of firms in country i expressed in currency j; pJit is the log of the aggregate price level of country i in currency j; p is the log of the steady state desired gross markup; and Г is the steady- state elasticity of that markup. Although strategic complementarities (Г) dampen their impact, the use of imported inputs in production has a direct effect on export and import prices via their effect on the marginal cost.
Note that these expressions are specific to the currency of invoicing, j. This feature has key implications for the relationship between the marginal cost of production and a firm's pricing decisions. When both exports and imports are priced in a common currency (the dominant currency), exchange rate fluctuations do not affect the terms of trade, but they do generate volatility in the relative prices of imported to domestic products. Hence, all else equal, the depreciation of a country's currency relative to the dominant currency, (1) reduces that country’s demand for imports from all countries, and (2) directly affects the real marginal cost of firms using imported inputs regardless of their country of origin.
Equations (6) and (7) are a good starting point to motivate the empirical exercises of this paper. They show that, in general, the use of imported inputs is directly related to changes in export and import prices. However, the inner workings of the model can be better illustrated in the case of fully flexible prices. Therefore, the remainder of this section will consider the scenario in which Sp = 0 so that the model yields the closed analytical forms used to generate the predictions that are tested with the data in Section 4.
Consider first the exchange rate pass-through into export prices. If domestic wages are rigid (Awt = 0) and productivity is unchanged (Aat = 0), the ERPT into export prices, expressed in H currency, is given by the following expression:
Under these restrictions, the pass-through boils down to the role of imported intermediate inputs, that affect the firms’ marginal cost, and the behavior of firms’ markups. If all intermediate inputs used in production are domestic (yH = 1) and markups are constant (Г = 0), the exchange rate pass-through into export prices is equal to zero (equivalently, there is full pass-through into destination currency prices), as firms set their local price as a constant markup above a fixed marginal cost regardless of the exchange rate. In contrast, when yH <1 and/or Г >0 the pass-through into export prices in home currency is positive (this is, there is an incomplete pass-through into destination currency prices).
The role of imported inputs is captured by the term. When
imported intermediate inputs are used in the production process, exchange rate fluctuations will have a direct impact on their cost. Therefore, a depreciation of the domestic currency will increase the marginal costs of production in H leading to an increase in local currency prices. Moreover, the effect is larger when intermediate inputs are relatively more important for production than labor. In other words, the marginal cost of production is decreasing in yH and increasing in a, such that the lower the home bias in intermediate inputs (the stronger the preferences for imported inputs) and the higher their share in production, the higher the ERPT into export prices. Importantly, although the presence of strategic complementarities can affect the impact of imported inputs on export prices, the positive correlation between the share of imported inputs and the ERPT into export prices holds for any given level of Г > 0.
Strategic complementarities and variable markups affect the ERPT in two ways. On one hand, there is positive and direct effect on ERPT, captured by the term j—pAeit. Firms do not want their prices to deviate from their competitors’ prices in foreign markets, so they keep prices in the destination country relatively stable and absorb exchange rate fluctuations by adjusting their markups and their prices in local currency. On the other hand, however, complementarities dampen the effect of the marginal cost channel because firms may not transmit all their cost shocks into prices in order to keep them stable and close to those of competitors.
Now, consider the exchange rate pass-through into import prices. Under the same assumptions, the ERPT into import prices in H currency is given by
for i, this case, strategic complementarities are crucial. From equation (9) it is clear that when markups are constant (Г = 0), the pass-through into home currency prices is equal to one regardless of the effect that exchange rate fluctuations have on domestic producers' costs through the use of imported inputs. With foreign firms setting a constant price in foreign currency, import prices in H adjust fully to the prevailing exchange rate. In contrast, with strategic complementarities (Г > 0), foreign firms set prices that depend on their local competitors' marginal costs, and the pass-through into import prices is incomplete.
Again, the effect of strategic complementarities and variable markups is twofold. The 1
last term, j+pdeit, captures the direct, negative impact of strategic complementarities in pricing; holding competitors' prices fixed, a higher Г reduces pass-through. But when foreign producers take into account the production costs of domestic firms, imported inputs affect import prices through an indirect effect that works in the opposite direction. Since a home currency depreciation increases marginal costs for H firms using of imported inputs, H firms will raise prices; in turn, it leads foreign firms to also raise theirs. This effect is increasing in Г and in the share of imported inputs in production, (1 — Yh ). In other words, in the presence of strategic complementarities (Г > 0), the ERPT into import prices is decreasing in Yh and increasing in a.
This discussion can be summarized in two testable predictions:
For my analysis, I use two firm-level data sets from Colombia: customs declarations that include information on several characteristics of firms' foreign transactions, and the national manufacturing survey that contains detailed operational information. The data used for the estimations presented in Section 4 cover the period between 2004 and 2015.
The data on firms' production and input consumption are from the Annual Manufacturing Survey, EAM (its acronym in Spanish), conducted by Colombia's department of statistics, DANE. The EAM is an unbalanced panel of Colombian factories with 10 or more employees, or sales above a certain threshold set every year. For each plant, the EAM report includes values for output and for several inputs used in the production process. Crucially for this study, it includes the share of imported intermediate inputs used at each plant. The data set also provides industry classification codes (under the Central Product Classification, CPC, and the International Standard Industrial Classification, ISIC) for the plants’ inputs and outputs.
The data on international trade, from Colombia’s customs agency (DIAN) and DANE, include information on the universe of Colombian exporters and importers. The data include each trading firm’s tax identification number; the 10-digit product code (according to the Nandina classification system, based on the Harmonized System); the FOB value (in U.S. dollars) and volume (net kilograms) of exports (imports); and the country of destination (origin), among other details. The data are monthly; for the empirical exercises, I aggregate exports and imports at the annual level.
I complement these data sets with macroeconomic information from the World Bank’s World Development Indicators. For the pass-through estimations, I use the annual average value of the official local currency unit per USD exchange rates (series PA_NUS_FCRF). Other controls include country-specific producer prices, measured with the wholesale price index (FP_WPI_TOTL), and real GDP, measured with the GDP at constant (2010 USD) market prices (NY_GDP_MKTP_KD). Colombian producer prices are measured with the manufacturing annual producer price index, calculated by DANE.
One important remark is that the two firm-level datasets cannot be merged; in order to keep information anonymous, the EAM identifies firms and plants with internal codes that cannot be matched to the firms' tax identification numbers. Hence, the analysis of the relationship between the use of imported inputs (observed in the EAM) and passthrough estimates (obtained with customs data) must be performed at an aggregate level. To include as many sectors as possible, results are presented at the 2-digit ISIC industry level. One advantage of this approach is that it can be used to study the relationship between ERPT and various industry characteristics, even when detailed micro data are not available.
I used the data to test the main implications of the model, stated in Section 2. Throughout the paper, I focus on manufacturing firms producing non-commodity, tradable goods. For the empirical analysis I exclude manufacturers of coke, refined petroleum products and nuclear fuel (ISIC 23), and manufacturers of basic metals (ISIC 27), which include metals such as gold, silver, platinum, and nickel, since as commodity producers their dynamics are different from those of other manufacturing firms. As a robustness check, I also limit the sample for some estimations to differentiated products, selected following the product classification proposed by Rauch (1999). Results remain qualitatively similar and are shown in the Appendix.
5.1. Industry Characteristics
I begin by documenting the heterogeneity in the use of imported intermediate inputs across manufacturing sectors. Differences across industries can stem from two sources: the share of domestic and imported inputs demanded by firms, and differences in the relative importance of intermediate inputs and labor in production (i.e., the a in the production function). I take advantage of the detailed information on input use contained in the EAM to analyze these two sources separately.
On average, imported inputs account for roughly one-fourth of the total expenditure on intermediate inputs purchased by manufacturing firms that are either exporters and/or importers (Table 2). Besides, this share is very stable across years. However, there is great heterogeneity across sectors. For instance, manufacturers of apparel (ISIC 18), leather products (ISIC 19), and wood products (ISIC 20) import around 10% of their inputs, while manufacturers of chemicals (ISIC 24), rubber and plastic products (ISIC 25), and different types of machinery and equipment (ISIC 30-35) rely more heavily on imported inputs. In addition, while the share of imported inputs is fairly stable across time for most industries and for the manufacturing sector as a whole, it fluctuates significantly for a few sectors (e.g., manufacturers of radio, television and communication equipment—ISIC 20 and ISIC 32).
Regarding the estimated share of inputs in production, I follow the method proposed by Gandhi, Navarro and Rivers (2017) and estimate a production function of the form Yt = eatXtL% for each 2-digit industry and for the manufacturing sector as a whole. I present the estimated values for the share of intermediate inputs (n) and the corresponding standard errors in Table 3. On average, the relative share of intermediate inputs in manufacturing production is estimated to be 0.45. Although there is some variation in the estimates, most sectors are within a standard deviation of the average value, and all sectors are within two.
A third variable that can affect the transmission of exchange rate fluctuations into trade prices, and that can be directly measured in the data, is the currency in which transactions are made. The distribution of the invoicing currencies for all exports between 2007 and 2014 shows that the U.S. dollar is overwhelmingly dominant for all sectors (Table 4). The shown distribution is calculated using the number of transactions, but the shares by export value are very similar. Given that the dollar dominance is common across industries and there is only slight degree of variation in invoicing currencies, I will focus on the use of imported inputs.
5.2. Exchange Rate Pass-Through into Prices
Given the observed heterogeneity in the use of imported intermediate inputs across manufacturing firms, the next step in my analysis is to estimate the exchange rate passthrough into import and export prices for each individual industry, and to check whether the estimates vary with these variables.
To estimate the pass-through of exchange rates into export and import prices, I regress the log change of unit values (my proxy for prices) on the log change of the annual average exchange rate, plus additional controls. Unit values are given by the FOB value per net kilogram, and are calculated at the 10-digit product/firm/country of destination (origin) level. In all cases, I cluster standard errors by year.
For transactions with the United States and other trade partners that use the US dollar as their currency, I estimate where P[j t denote export (import) prices measured in Colombian pesos; £ut is the average nominal exchange rate of the Colombian peso relative to the dollar; and Xt includes price indices for Colombia and the corresponding country of destination (origin) and fixed effects by firm-industry-country. This empirical approximation to the model presented in Section 2 implies that for export prices, and for import prices. In both cases the model predicts that the estimates for p should be increasing with the use of imported inputs under reasonable Y parameter values.
Colombian peso relative to the dollar. Here, for the export price estimations, and fiB = -+p[l + Г ] and ри =for the import price estimations. Adding an additional exchange rate to the estimations allows testing the relative importance of the local currencies and the dominant dollar in the pass-through. It also allows a check on whether the share of imported inputs and/or the variability of markups are more strongly related to one exchange rate or the other.
The estimates for the manufacturing sector as a whole and for each 2-digit industry are in Tables 5-10. As a robustness check, the estimations were also made only for differentiated products. The estimates are in Tables A5-A10 in the Appendix.
The estimated ERPT into export prices for dollarized destinations differs greatly across industries (Table 5). In general, sectors that use a larger proportion of imported inputs have larger ERPT estimates than those that use mostly domestic inputs. For instance, manufacturers of apparel, leather, and wood products (ISIC 18-20) use the fewest imported inputs and, accordingly, have the lowest ERPT into export prices. ERPT estimates are significantly higher for producers of chemicals (ISIC 24) and of rubber and plastic products (ISIC 25), who rely heavily on imported inputs.
The relationship between the ERPT estimates and the share of inputs in the production function is less clear, perhaps because of the lack of variation in the estimated values of a (Table 3). Notably, however, the estimated ERPT for the food industry (ISIC 15) is among the highest. Firms in this sector use a smaller share of imported inputs than the average manufacturer, but intermediate inputs are more important for the productive process of this sector than for any other. These results suggest that the marginal cost channel is important in explaining pass-through differences across exporting firms.
Estimates for the ERPT into export prices for other destinations include the results of two regressions for each industry: (1) the pass-through of the corresponding bilateral exchange rate and (2) the first regression with the peso/dollar exchange rate as an additional regressor. This procedure allows me to assess the relative importance of the two exchange rates for export prices. It also allows a check on how the use of imported inputs is related to each exchange rate.
When only the bilateral exchange rate is included (Tables 6-7, odd-numbered columns), the estimated ERPT values are still correlated with the intensity of use of imported inputs, but the relationship is not as clear as in the case of the dollarized economies. Producers of chemicals (ISIC 24), electrical machinery (ISIC 31) and vehicles and transport equipment (ISIC 34-35), all heavy users of imported inputs, have the largest pass-through. But the relationship is not clear for other sectors. The ERPT estimates for the bilateral exchange rate have a correlation coefficient of 0.10 for the average share of imported inputs, and 0.06 for the share of intermediate inputs in production.
Once the peso/dollar exchange rate is included (Tables 6-7, even-numbered columns), it greatly reduces the effect of the bilateral exchange rate and, in most cases, it eliminates its effect completely. The estimated ERPT of the peso/dollar rate is more homogeneous across sectors, but estimates are still positively correlated with the use of imported inputs. The correlation in this case is 0.32, whereas the correlation with the estimates for the bilateral rate is -0.01. The correlation between both sets of estimates and the values for a is also low.
The estimated coefficients for the ERPT into import prices from dollarized economies (Table 8) are very similar across sectors (with the notable exception of ISIC 18), and the value does not seem to vary much with the share of imported inputs presented in Table 2. Excluding ISIC 18 from the calculations, the correlation between ERPT estimates and the average share of imported inputs is 0.11, much lower than in the case of export prices. This makes sense because strategic complementarities dampen the effect of the marginal cost channel on import prices. As illustrated by the model, the impact of imported inputs on the price of imports is indirect, affecting them only through the sellers’ incentive to keep them close to those of local competitors.
Estimates for the ERPT into import prices from non-dollarized economies consist again of two sets: bilateral exchange rates only, and bilateral plus the peso/dollar rate (Tables 9-10). When only the bilateral rates are included, the coefficients do not vary much the share of inputs in production. As in the case of export prices, adding the peso/dollar exchange rate results in much lower coefficients. Moreover, with the addition of the peso/dollar rate, the share of imported inputs is positively correlated with the estimated pass-through of the peso/dollar rate (the correlation is 0.46) but not with the bilateral rate (-0.09). Neither set of estimates exhibits a clear correlation with the values for a.
5.3. Exchange Rate Pass-Through into Quantities
Besides being informative about pricing behavior, the analysis of the ERPT is important to understand the extent of expenditure switching that can follow exchange rate fluctuations. These potential changes in quantities are what ultimately matter for welfare. Given the substantial heterogeneity across industries in the ERPT into prices, the question that follows naturally is whether the ERPT into traded quantities is also heterogeneous.
The DCP model described in Section 2 predicts that the ERPT into exported quantities should be fairly small (if not entirely negligible), while imported quantities should be much more responsive to exchange rate variations. After taking a first order approximation, the ERPT to quantities exported (yHi) and imported (уш) is given by
It follows that the ERPT into quantities exported to dollarized economies (country U) is predicted to be around zero since = 1 and is close to 1. In the case of the exports to the rest of the world (country R), the effect will also depend on the comovement between the exchange rates of U and R, ^-. Intuitively, if both exchange rates are highly correlated, the change in the quantities exported to R will be mostly flat (as in the case of U); however, if the correlation is lower, a home currency depreciation relative to the dominant currency will lead to a decline in the exports to R because of its now weaker currency.
In the case of the ERPT into imported quantities, given that is also close to 1, it follows that, after controlling for demand (yd) and home competitor prices (p), the quantities are almost as sensitive as the elasticity of demand (a). Furthermore, in all cases, the only exchange rate that matters is the one with respect to the dominant currency (еи) and not the bilateral rate (eR).
To test the sensitivity of exported and imported quantities to changes in exchange rates, the log change of the volume of exports (imports) is regressed on the log change of the annual average exchange rate and on additional controls. Quantities are measured in net kilograms and, again, observations are at the 10-digit product/firm/country of destination (origin) level. As in the case of prices, I estimate the pass-through separately for dollarized and non-dollarized economies. For transactions with the United States and other trade partners that use the US dollar as their currency, I estimate
where £ut is the average nominal exchange rate of the Colombian peso relative to the dollar; and Xt includes price indices for Colombia and the corresponding country of destination (origin), the buyer’s GDP growth to account for demand effects, and fixed effects by firm-industry-country. For other trading partners, I estimate
where £Bt is the average nominal bilateral exchange rate of the Colombian peso relative to the local currency unit, and £ut is the average nominal exchange rate of the Colombian peso relative to the dollar. All standard errors are clustered by year.
The estimates are presented in Tables 11-16. Each set of destinations has estimates for the manufacturing sector as a whole and for each 2-digit industry. Again, the Appendix contains results for the subset of differentiated goods (Tables A-11-A-16).
As predicted by the model, the response of exported quantities to dollarized economies is generally relatively small, and roughly half of the estimated coefficients are statistically insignificant (Table 11). Moreover, as in the case or ERPT into prices, the response of quantities to exchange rate changes differs greatly across sectors, and the variation in industry-specific estimates is consistent with the predictions of the model. Some of the industries with a relatively high share of imported inputs (e.g., ISIC 24, 32) have high ERPT into prices that also results in a relatively large decrease in quantities. Sectors that rely more on domestic inputs and therefore have a lower ERPT into prices (e.g., ISIC 18, 19) do not adjust their exported quantities significantly. The ERPT estimates have a correlation coefficient of -0.36 for the share of imported inputs and - 0.27 for the share of inputs in production.
The response of quantities for non-dollarized destinations is negligible for several industries (Tables 12-13). When quantities do respond, the relevant exchange rate is the peso/dollar rate. However, the relationship between the estimates and the use of imported inputs in production is less clear in this case; among the industries with the largest ERPT estimates are the two sectors that rely mostly on domestic inputs (ISIC 18 and 19) and sectors that use a high share of imported inputs (ISIC 24 and 34+35). In fact, contrary to the prediction of the model, the correlation between the two variables is positive.
All import quantities, regardless of their country of origin, respond to changes in the peso/dollar exchange rate (Tables 14-16). For the aggregated manufacturing sector and for most individual industries, a peso depreciation is associated with a decrease in imports. Although the magnitude of the response varies somewhat across industries, no clear relationship appears between the differences in ERPT estimates and the analyzed industry characteristics.
This paper evaluates whether there is any heterogeneity in the impact of exchange rate fluctuations across manufacturing industries.
The paper first examines the use of imported intermediate inputs in production, a key feature known to affect the exchange rate pass-through, and finds that there are significant variations across industries. The heterogeneity can be explained mainly by differences in the share of imported inputs in firms' total input purchases; some sectors rely much more than others on imported inputs. Industries are also heterogeneous in the relative importance of intermediate inputs in production, although the differences in terms of technology are smaller.
The paper then estimates the exchange rate pass-through (ERPT) into export and import prices for the manufacturing sector as a whole and at the 2-digit industry level. It finds that the response of prices to exchange rate fluctuations varies greatly across industries. Despite these differences, the ERPT into trade prices is generally high (close to 1) for both exports and imports, and the most relevant exchange rate is the peso/dollar rate regardless of the destination or origin of trade.
Connecting the information on input use and the ERPT estimates at the industry level provides evidence of the effect of the marginal cost channel on the ERPT. It works mainly through the changes in cost that result from using imported inputs. Industries with a higher share of imported inputs have a higher pass-through. As expected, the effect seems to be stronger in the case of export prices, while the effect on import prices is moderated by the presence of strategic complementarities.
Finally, the paper estimates the exchange rate pass-through into exported and imported quantities. Changes in the peso/dollar exchange rate have a relatively small effect on Colombian exports, even negligible for some industries, but have a significant effect on all imports regardless of origin. Although the size of the estimated ERPT into quantities differs across industries, no clear pattern relates these estimates to the use of imported inputs.
Amiti, M., Itskhoki, O., and Konings, J. (2014): “Importers, Exporters, and Exchange Rate Disconnect,” American Economic Review, 104(7), 1942-1978.
Amiti, M., Itskhoki, O., and Konings, J. (2018): “International Shocks, Variable Markups and Domestic Prices,” Forthcoming in the Review of Economic Studies.
Atkeson Andrew, and Burstein, A. (2008): “Trade Costs, Pricing-to-Market, and International Relative Prices,” American Economic Review, 98(5), 1998-2031.
Auer, R. A., and Schoenle, R. S. (2016): “Market structure and exchange rate passthrough,” Journal of International Economics, 98, 60-77.
Berman, N., Martin, P., and Mayer, T. (2012): “How Do Different Exporters React to Exchange Rate Changes?,” Quarterly Journal of Economics, 127(1), 437-492.
Burstein, A., and Gopinath, G. (2014: “International Prices and Exchange Rates,” in Handbook of International Economics, ed. by G. Gopinath, E. Helpman, and K. Rogoff, Vol. 4, pp. 391-451. Elsevier.
Burstein, A., Neves, J., and Rebelo, S. (2003): “Distribution costs and real exchange dynamics?,” Journal of Monetary Economics, 50, 1189-1214.
Bussiere, M., Delle Chiaie, S., and Peltonen, T. A. (2014): “Exchange Rate Pass-Through in the Global Economy: The Role of Emerging Market Economies,” IMF Economic Review, 62(1), 146-178.
Campa, J. M., and Goldberg, L. S. (2005): “Exchange Rate Pass-Through into Import Prices,” The Review of Economics and Statistics, 87(4), 679-690.
Casas, C., Diez, F., Gopinath, G., and Gourinchas, P.-O. (2017): “Dominant Currency Paradigm,” IMF Working Paper No. 17/264.
Chen, N., and Juvenal, L. (2016): “Quality, trade, and exchange rate pass-through,” Journal of International Economics, 100, 61-80.
Fauceglia, D., Shingal, A., and Wermelinger, M. (2014): “Natural Hedging of Exchange Rate Risk: The Role of Imported Input Prices,” Swiss Journal of Economics and Statistics, 150(4), 261-296.
Fitzgerald, D., and Haller, S. (2014): “Pricing-to-Market: Evidence From Plant-Level Prices,” Review of Economic Studies, 81 (2), 761-786.
Gandhi, A., Navarro, S., and Rivers, D. (2017): “On the Identification of Gross Output Production Functions,” Working Paper.
Garetto, S. (2016): “Firms' heterogeneity, incomplete information, and pass-through,” Journal of International Economics, 101, 168-179.
Goldberg, P. K., and Hellerstein, R. (2008): “A Structural Approach to Explaining Incomplete Exchange-Rate Pass-Through and Pricing-to-Market,” American Economic Review, 98(2), 423-429.
Gopinath, G., Itskhoki, O., and Rigobon, R. (2010): “Currency Choice and Exchange Rate Pass-Through,” American Economic Review, 100(1), 304-336.
Greenaway, D., Kneller, R., and Zhang, X. (2010): “The Effect of Exchange Rates on Firm Exports: The Role of Imported Intermediate Inputs,” The World Economy, 33(8), 961-986.
Rincon-Castro, H., and Rodriguez-Nino, N. (2018): “Nonlinear state and shock dependence of exchange rate pass-through on prices,” BIS Working Paper No. 690.