Exchange Rate, External Orientation of Firms and Wage Adjustment

Exchange Rate, External Orientation ofFirms and Wage Adjustment

Francesco Nucci1 and Alberto Franco Pozzolo21Sapienza University, Rome, Italy and 2Universita’ del Molise, Campobasso, Italy

1. INTRODUCTION

A considerable body of empirical literature in international economics uncovers non-negli-

gible effects of exchange rate movements on the real economy. For example, Campa and

Goldberg (1999) and Nucci and Pozzolo (2001) show that currency swings significantly affect

firm’s investment, and Nucci and Pozzolo (2010) and Goldberg and Tracy (2000) find similar

effects on employment and hours worked.1 From a theoretical perspective, the typical channel

through which exchange rates can affect firms’ choices is through changes in profitability.

A currency depreciation, for example, causes an increase in the costs of imported inputs for

production, on one side, and an increase in foreign and domestic revenues on the other. The

overall impact is therefore expected to vary considerably across producers in the light of the

heterogeneous international orientation on both the export and imported inputs side of each

individual firm (see Bernard et al., 2007).

A parallel strand of empirical literature has studied the consequences of exchange rate

movements on the earnings of workers, but without reaching a definitive conclusion. On the

one hand, a group of contributions, focusing on aggregate data at the industry level, uncovers

a significant wage responsiveness to exchange rate oscillations. In particular, Revenga (1992)

shows that changes in import prices induced by exchange rate swings do affect real wage in

US manufacturing, Goldberg and Tracy (2000), using data disaggregated by states and by

industries, report statistically significant estimates of the earnings elasticities with respect to

exchange rate, also providing evidence of a pattern of regional differences in their values,

and Campa and Goldberg (2001) document a significant incidence of exchange rates on

wages, which is shown to vary across industries depending on their trade orientation and

competitive structure. On the contrary, Goldberg and Tracy (2001), using information on

individual workers characteristics such as educational attainment and experience, find that the

overall wage effect of currency movements is in fact modest, although it can be sizeable for

specific groups of workers.

In this paper, we contribute to the literature on the effect of exchange rate fluctuations on

wages relying on microeconomic data from a high-quality sample of Italian manufacturing

This project was begun while Alberto Pozzolo was with the Research Department of the Bank of Italy.We are especially grateful to two anonymous referees for their helpful comments, and we also thankLinda Goldberg, Paolo Naticchioni and seminar participants at the IEA Conference in Beijing, theUniversity of Trento and the XIX Tor Vergata University Conference on Money, Banking and Finance(Rome) for useful discussions.

1 The latter finding contrasts with previous evidence by Campa and Goldberg (2001) of a weak employ-ment response to exchange rates coupled with a significant impact on real wages; unlike Campa andGoldberg (2001) and most other studies in this area of literature, Nucci and Pozzolo (2010) use firm-rather than industry-level information and provide arguments for why the level of disaggregation of thedata may affect the estimated impact on employment to a considerable extent.

© 2014 John Wiley & Sons Ltd 1589

The World Economy (2014)doi: 10.1111/twec.12200

The World Economy

firms, thus complementing the Nucci and Pozzolo’s (2010) analysis along a parallel

dimension.2 Information on individual workers, as in Goldberg and Tracy (2001), definitely

provides an important perspective, as it allows to appraise which type of workers experience

the largest wage adjustment (if any) after an exchange rate shock.3 We argue, however, that

individual firm characteristics are of particular relevance in shaping the implications for wages

of exchange rate fluctuations. In particular, because of the large heterogeneity across firms

along several dimensions, even within narrowly defined industries, the firm-specific effects of

exchange rate on wages can be to some extent washed out in the aggregation process. This

paper therefore analyses if the estimated wage response to an exchange rate shock varies

across firms in association with the response of firm profitability.

In frictionless and competitive labour markets, the change in profitability resulting from an

exchange rate swing is not expected to yield any firm-specific effect on the wages of similar

workers, as each firm is a wage-taker that faces an infinitely elastic supply of labour. Any

shift in the marginal revenue product of labour induced by a currency shock would generate

firm-level effects on employment, but not on wages. However, this contrasts with the well-

established empirical finding that firm profitability does affect its wages, even after controlling

for differences in the characteristics of the workers (see e.g. Nickell and Wadhwani, 1990;

Abowd and Lemieux, 1993; Arai, 2003). Indeed, a variety of theoretical models depart from

the simplest auction model of the labour market and yield the prediction that workers are paid

a wage that does depend on the profits and the financial conditions of their employer.4

Thus, the combined evidence that the effects of currency swings on profitability and labour

demand are heterogeneous across firms, depending on their international exposure, and that

differences in profitability across firms are conducive to differences in wages of similar work-

ers calls for an investigation on firm-level data of the exchange rate-wages link. With our

data, we are able to estimate a specific, time-varying responsiveness of the (average) wage to

exchange rate for each individual firm, and in doing so, we explicitly allow for a number of

transmission channels of the currency shock to the wage rate whose relevance varies from

firm to firm.

The theoretical motivation for the empirical analysis stems from the formal models devel-

oped in Campa and Goldberg (1999; 2001) and Nucci and Pozzolo (2001; 2010), which pin

down a number of mechanisms through which exchange rate changes may influence real vari-

2 Our data span manufacturing firms only and exclude firms from other sectors which are also exposedto foreign markets; hence, being specific to a portion of the economy, our findings cannot be generalisedto all sectors.3 Indeed, two further advantages of individual data are that they allow to appraise the effect of exchangerate on the wage of workers that decide or are forced to change their job (see Kletzer, 2000; Goldbergand Tracy, 2001) and to investigate how exchange rate affects wage inequality.4 As elucidated by Blanchflower et al. (1996) and Hildreth and Oswald (1997), for example, an impor-tant class of models features a bargaining framework where rent-sharing occurs between the firm and itsemployees, and this establishes a positive link at the firm level between profitability and pay. Alterna-tively, a competitive model with temporary informational and search-related frictions implies that theindividual firm faces an upward-sloping labour supply curve, so that a shock to profitability, induced forexample by an exchange rate shock, feeds through into the wage rate, not just into employment. Modelswhere labour contracts have elements of risk-sharing are also conducive to a positive response of thefirm wage rate to profits, whose size depends on the degree of relative risk aversion of the workers andthe firm. Vandenbussche and Konings (1998) present a related theoretical and empirical analysis ofthe impact of import competition on employment and wages when labour markets are imperfectlycompetitive.

© 2014 John Wiley & Sons Ltd

1590 F. NUCCI AND A. F. POZZOLO

ables chosen by the firm.5 The prediction that exchange rates affect not only a firm’s labour

demand but also the level of its wages is obtained by assuming that in the short run labour is

imperfectly mobile, due to search and informational frictions that make each employer facing

an upward-sloping labour supply curve, rather than a horizontal one at the economy-wide

wage, as if the labour market were perfectly competitive. In this setting, a currency deprecia-

tion causes a surge in equilibrium profitability and in the wage level, through the revenue side

of the income account, and the effect is stronger the higher the incidence of a firm’s sales in

the foreign markets. Conversely, through the cost side, a depreciation causes a reduction in

profits and wages, and the effect in turn is stronger the more a firm relies on imported inter-

mediate inputs. Another theoretical prediction, also testable with our data, is that, for a given

level of firm’s external orientation, the degree of wage sensitivity to exchange rates is influ-

enced by the market power of the firm. In particular, the impact on wages can be shown to

be more pronounced the lower is the firm’s pricing power in its destination markets (see

Campa and Goldberg, 2001).6

Hence, our goal in the empirical analysis, rather than uncovering the aggregate, average

effect of currency swings on wages, is to assess the degree of specificity across firms in the

wage sensitivity to exchange rate swings that arise from differences in their international ori-

entation. This degree of specificity arises as a key prediction of our theoretical framework

and originates primarily from differences across firms in their external orientation on both the

export and imported inputs side. Against this background, the Italian manufacturing sector

over the period antecedent to the advent of the euro (1984–98) provides an interesting and rel-

evant case study, since it refers to an industrialised economy, highly exposed to international

trade and with the exchange rate exhibiting pronounced swings over the sample considered.7

The empirical analysis soundly supports the theoretical predictions. An exchange rate

devaluation causes an increase in wages that is stronger the higher is the share of revenues

from exports and the lower is the incidence of imported inputs, and such effects are magnified

the lower the firm’s market power. We also investigate the empirical relevance of additional

features that shape the responsiveness of firm-level wages to exchange rate fluctuations, in

particular the effect of the competitive pressure exerted by foreign producers in the firm’s

domestic market. Following Nucci and Pozzolo (2010), we investigate whether differences

across industries in the degree of import penetration are conducive to diverging patterns in

the response of wages to currency shocks. In doing so, we combine firm-level information on

the exposure of the firm’s sales on the domestic product markets with data on import penetra-

tion in the industry to which the firm belongs. We find that a currency depreciation, by reduc-

ing the competitiveness of importers in the domestic market, increases the profitability of

domestic firms and thereby their wages, through the domestic revenue channel. This effect is

5 See also the working paper version of the present paper, Nucci and Pozzolo (2012).6 Along a similar line of reasoning, Greenaway et al. (2010) show that exchange rates have no signifi-cant effect on firm exports because the negative effect of an appreciation due to the lower competitive-ness on the revenue side is offset by the increased competitiveness coming from a lower cost ofimported intermediate inputs.7 The subsequent introduction of the euro in 1999, on the other hand, disconnects to some extent theexternal orientation of firms from their exchange rate exposure as a notable part of export sales and pur-chases of imported inputs arises within the common currency area. On this respect, this would reducethe heterogeneity across firms in the wage responsiveness to exchange rate shocks and, importantly toour purposes, weaken the link characterising our theoretical framework between the firm’s openness tointernational trade and its exchange rate exposure.


EXCHANGE RATE AND WAGE ADJUSTMENT 1591

found to be stronger the higher is the extent of import penetration and the larger is the degree

of firm’s orientation to the domestic product market.

We organise the remainder of the paper as follows: Section 2 discusses the theoretical pre-

diction from previous contributions that guide our empirical analysis and provides motivation

for it. Section 3 describes the data, the empirical specification and the estimation methodol-

ogy. In Section 4, we report the empirical results with regard to the baseline specification as

well as to additional features characterising the wage responsiveness to exchange rate swings.

Section 5 concludes.

2. THE CHANNELS OF TRANSMISSION OF THE CURRENCY SWINGS TO WAGES

By considering the optimal conditions for profit maximisation of a firm operating in imper-

fectly competitive product markets, Campa and Goldberg (1999, 2001) and Nucci and

Pozzolo (2001, 2010) propose a theoretical framework that identifies the channels of transmis-

sion of exchange rate movements to real variables set by the firm. This framework can be

used for analysing how exchange rate swings induce equilibrium wage adjustments.

When dealing with this issue, it is important to clarify that if the firm were a wage-taker

and faced an infinitely elastic supply of labour at the economy-wide prevailing wage level,

then an exchange rate shock would yield a shift in profitability and labour demand, which

would impact on the firm’s equilibrium employment but would leave the wage rate unaltered.

Since this contrasts with the well-established evidence that firm profitability is a relevant

determinant of workers pay, even when controlling for their observable differences, we depart

from the assumption of a perfectly competitive labour market and consider one of the several

frameworks in which firm wages and profits exhibit comovements. The latter arise, for exam-

ple, if a modified competitive model of the labour market is considered, in which the labour

supply curve slopes upwards. This may originate from temporary frictions resulting from

search and mobility costs, informational problems or heterogeneous worker preferences, as

suggested for example by Blanchflower et al. (1996) and Hildreth and Oswald (1997).8

In accordance with Campa and Goldberg (1999, 2001) and Nucci and Pozzolo (2001,

2010), manipulating the firm’s optimal conditions and combining them with a labour supply

relationship yields an expression for the elasticity of real wages, w, with respect to the

exchange rate, e, that identifies the different channels through which exchange rate move-

ments affect the equilibrium outcome for the firm wages.

If we focus on the revenue side of the firm’s balance sheet, a currency depreciation has a

positive effect on the firm’s marginal revenue product of labour and thereby on the equilib-

rium wage. In particular, a depreciation is predicted to exert a positive effect on wages along

both the foreign and domestic sales channels. On the export side, the positive effect on wages

of an exchange rate depreciation is larger the higher the share of foreign sales in total sales,

that is, the more a firm is exposed to foreign markets through the export of its products. On

the revenue side as a whole, that is, including both domestic and foreign sales, the positive

effect on wages of a currency depreciation is also amplified as the export share of sales

becomes larger, but this prediction holds true only if the sum of the exchange rate

8 Albeit implicitly, the same hypothesis on the labour market is adopted by Campa and Goldberg(2001), where wages are allowed to be different across industries in the light of a number of sector-spe-cific characteristics.



pass-through elasticities (their absolute value) is less than one.9 By contrast, through the cost

side of the balance sheet, a currency depreciation negatively affects the marginal revenue

product of labour and thereby wages. Again, the extent of these adjustments depends on the

external dependence of the firm on foreign inputs. This is captured by the share of expendi-

ture for imported inputs in total costs: the larger this share, the more pronounced are the neg-

ative implications for wages of a depreciation through this channel of exposure (see Campa

and Goldberg, 1999, 2001; Nucci and Pozzolo, 2001, 2010).

The theoretical framework devised in previous contributions highlights other important fea-

tures that shape the response of wages to exchange rate swings. First, the degree of firm’s

market power, as measured by the firm’s mark-up ratio, with the prediction that, everything

else being identical for two hypothetical firms (e.g. the type of foreign exposure and the pass-

through elasticities), the one with a lower degree of pricing power exhibits a more sizeable

elasticity of wages to exchange rate (in absolute value). Indeed, firms with a lower market

power tend to be less capable of absorbing currency shocks, so that the impact on wages is

relatively more pronounced. The intuition is that an exchange rate depreciation drives down

the export price in the foreign currency by an amount that depends on the pass-through elas-

ticity. This price decline therefore yields an increase in foreign demand – and thereby of prof-

itability and wages – which is larger the higher is the price elasticity of foreign demand.

Given the negative relationship between the firm’s mark-up ratio and the price elasticity of

demand, the sensitivity of wages to currency swings is indeed magnified when the firm’s

market power is relatively low.

Another theoretical prediction is that the exchange rate pass-through elasticities contribute to

shape the adjustment of wages in response to currency shocks. The exchange rate elasticity of

the prices set by the firm in the currency of the market of destination ranges from zero (no pass-

through) to one (complete pass-through). For a given level of external orientation on the export

side, the smaller is the (absolute value of the) elasticity of exchange rate pass-through to foreign

prices, the larger is the wage response to a shift in the exchange rate (see Nucci and Pozzolo

2010). Indeed, many contributions show that this exchange rate pass-through elasticity does

depend on market structure and in particular on the extent to which firms’ products are differen-

tiated and the substitution among different variants is large (Yang, 1997). These studies, which

include Dornbusch (1987) and Knetter (1993), show that the pass-through tends to be low if the

degree of competition in the foreign markets is high. Therefore, in the limiting case of a per-

fectly competitive foreign destination market, the firm is a price-taker, the pass-through elastic-

ity is zero, and the effect of a depreciation on profits and wages is magnified. This additional

channel clearly reinforces the previous conclusion that the lower the firm’s pricing power, the

higher the exchange rate sensitivity of profitability and thereby wages.

If we focus on firm’s competition in the domestic market, we have already recalled that a cur-

rency depreciation, by making foreign products more expensive, rises the competitiveness of

domestic firms in the home market, thus increasing their sales, their profitability as well as their

wages. Moreover, the literature predicts also that the exchange rate pass-through elasticity of

firm’s prices in the domestic market plays an important role in shaping this effect. The value of

this domestic pass-through elasticity ranges from minus one (complete pass-through) to zero (no

pass-through), and again it depends on market structure. Specifically, the elasticity (in absolute

9 See Nucci and Pozzolo (2010) who recall that the exchange rate pass-through elasticities are the elas-ticities of the domestic and of the foreign prices with respect to the exchange rate.



value) is a decreasing function of the firm’s monopoly power in the home market. In the limiting

case of a perfectly competitive domestic destination market, where the domestic firm is a price-

taker, a currency appreciation forces it to a one-for-one price cut (a pass-through elasticity equal

to minus one). Thus, a currency appreciation lowers the value of domestic sales, profitability and

wages, and the lower the firm’s market power, the larger is this decline. The intuition is that the

higher the competitive pressure exerted by foreign producers, the more responsive are domestic

sales, profitability and wages in the aftermath of a currency shift. Indeed, the domestic pass-

through elasticity reflects the degree of this competitive pressure from foreign producers and is

often assumed to be proportional to the degree of import penetration in the domestic market (see

e.g. Dornbusch, 1987; Campa and Goldberg, 2001).

Market structure also shapes the effects of exchange rate fluctuations through the cost side

of the balance sheet. The increase in wages caused by an exchange rate appreciation through

a reduction in the expenditure for imported inputs depends on the extent of competition in the

market for intermediate inputs. The effect of exchange rate swings on wages is larger, the

smaller the pass-through elasticity of foreign input prices to the exchange rate that ranges

from zero (no pass-through) to one (complete pass-through).

This discussion, based on the theoretical background devised in previous contributions to the

literature (see Campa and Goldberg, 1999, 2001; Nucci and Pozzolo, 2001, 2010), has allowed

us to uncover a number of testable implications on the transmission of exchange rate swings to

wages. These theoretical predictions lend themselves to the empirical scrutiny, to which we now

turn. In doing so, we first present the firm-level panel data that we use in our analysis.

3. THE DATA AND THE REGRESSION SPECIFICATION

a. The Data

The microeconomic data used in the empirical investigation are the the same as those of

Nucci and Pozzolo (2010). They are drawn from two different statistical sources. The first

one is the Bank of Italy’s Survey of Investment in Italian Manufacturing (SIM) that has been

carried out at the beginning of every year since 1984 on a representative sample of over

1,000 manufacturing firms, stratified by industry, size and location.10 These data are of extre-

mely high quality, thanks also to the professional expertise of the interviewers, who are offi-

cials of the Bank of Italy establishing long-term relationships with the firms’ managers. Only

medium–large firms, defined as those with more than 50 employees, were included in the sur-

vey during the sample period that we analyse, accounting for more than 40 per cent of total

workforce in the Italian manufacturing sector.11

We used SIM to gather firm-level information on total revenues and revenues from export-

ing, employment and hours worked. The second source of firm-level data is the Company

10 Since the effects of exchange rate fluctuations on wages that we seek to investigate work primarilythrough the firms’ external exposure on foreign markets, they are especially relevant for firms producingtradeable products. This validates our focus on manufacturing firms that in Italy accounted for about23 per cent of total employment in the period 1984–98, with a slight reduction to 20 per cent in theperiod thereafter.11 Given that the status of exporter for a firm is typically associated with larger size (see e.g. Bernardet al., 2007), not having small firms in the sample is not problematic in our analysis because we focuson the firm’s degree of external orientation – that is, the intensive margin of trade – as the key determi-nant of the responsiveness of the firm’s wages to currency shocks.



Accounts Data Service reports, a database maintained by a consortium among the Bank of

Italy and a pool of commercial banks, collecting information from balance sheets and income

statements of a sample of about 40,000 Italian firms. The detailed information from the

annual balance sheets are reclassified to ensure comparability across firms. From this data-

base, we gather firm-level information on the wage bill, expenditure on intermediate inputs,

value added, gross output and total sales. Data from these two sources are merged to construct

an unbalanced panel of slightly fewer than 2,400 firms. As in Nucci and Pozzolo (2010), the

data used for estimation cover the period 1984–98.The 15-year period ahead of the introduction of the euro provides a suitable case study for

assessing the heterogeneous responsiveness of firms’ wages to exchange rate shocks. During

these 15 years, the Italian currency lira exhibited pronounced swings, influenced mainly by

the bilateral exchange rates with respect to the German Mark and the other currencies taking

part to the European target zone system known as Exchange Rate Mechanism (ERM), and

with respect to the US$. Between 1984 and 1990, the lira had an oscillation band of �6.0

per cent within the ERM, while currencies of the other members had an oscillation band of

�2.25 per cent. The ERM allowed for agreed realignments among the member countries, and

this permitted the lira to absorb in part the appreciation caused by the higher inflation rate in

Italy. After the last realignment involving the lira, that took place in April 1986, the Italian

currency experienced a light depreciation in the subsequent two years. Starting from 1989, the

lira begun a period of steady appreciation. After it adhered to an oscillation band of �2.25

per cent in May 1990, the value of the Italian currency was sustained by significant capital

inflows, attracted by the higher expected stability of the exchange rate. However, following

the rejection of the Maastricht Treaty by Denmark, in June 1992, the ERM came under a

fierce attack that determined on 16 September the exit of the Italian lira and of the British

pound (that had joined the mechanism only in 1990). With the exit from the ERM, the lira

experienced immediately a strong drop in its value that amounted to about 30 per cent in the

following two years. However, the strong fiscal discipline enacted by the Government, cou-

pled by the strict monetary policy chosen by the Bank of Italy, with the aim of joining the

European Monetary Union from the beginning, determined in the following years a steady

appreciation of the lira which rejoined the ERM in November 1996 (Figure 1). Due to all

these events, the standard deviation of the Italian real effective exchange rates (based on unit

labour costs) was 9.8 in the sample 1984–98, while it dropped to 5.6 thereafter.

This marked variability of the Italian exchange rate in the pre-euro period, coupled with

the high degree of international exposure of Italian firms on both the revenue and the cost

side, provides an ideal setting for our empirical analysis.12 Moreover, addressing the issues of

this paper with data on Italian firms is appropriate for the characteristics of the labour market

in Italy. The latter was by no means frictionless and competitive, and therefore, it is reason-

12 It is important to notice that most contributions to the literature studying the effects of the introduc-tion of the euro find a positive but limited impact on manufacturing trade flows within the euro area,with estimates ranging between 2 and 5 per cent; the effect is even lower for international trade flowswith countries outside the euro area (see Banca d’Italia, 2009 and the reference therein). Clearly, whilethe extent of trade openness has not significantly changed as a result of the introduction of the euro, iffirms had increased their external orientation towards euro area countries, on both the export and importside, this would not have translated anymore into a higher exposure to exchange rate fluctuations. Onthe other hand, this lower dependence on exchange rate oscillations is mirrored by the lower variabilityof Italy’s real effective exchange rate since 1999.



able to believe that firms operating in it are far from being wage-takers. These firms do have

some power in setting workers’ wages as the theoretical framework underlying our empirical

analysis suggests. On the other hand, wage-setting institutions prevailing in Italy prevent

wages from being fully flexible. Indeed, wage-setting characteristics differ widely among

countries, and one important dimension is the degree of bargaining centralisation. Cross-

countries indicators of such centralisation have been computed by the OECD and, in the

sample period analysed in the paper, Italy is grouped among the countries with a ‘combina-

tion of industry and company/plant level of centralisation with an important share of employ-

ees covered by company bargains’ (OECD, 2004, Table 3.5). In countries such as the US,

UK, Japan, Canada and Korea, wage negotiations are predominantly at the firm level, while

Nordic countries represent the other extreme with a predominance of inter-industry bargaining

at the national level (see OECD, 2004).

We compute the average wage at the firm level as the total wage bill divided by the aver-

age number of employees within each year (average wage per employee) or by the total hours

worked (average wage per hour worked). To obtain measures of labour compensation in real

terms, we deflate them using the GDP deflator. We also have information on the incidence at

the firm level of white-collar (or non-production) employees. The latter include personnel

engaged in supervision, servicing, professional, technological and administrative. As is com-

mon in the literature, we use the incidence of white-collar workers to proxy for the skill

intensity of the workforce (see e.g. Berman et al., 1994; Voigtlnder, 2013).

The empirical counterpart of the two key variables on the firm’s international exposure is

calculated for each year as follows. The export share of sales of firm i in each year t, vit, isdirectly provided by SIM. As for the import share of input expenditure, unfortunately it is not

possible to distinguish between domestically produced and imported inputs from the income

statements of each firm. In order to deal with this limitation of our data and provide an

approximation for the share of imported inputs in total input expenditure of firm i in year t,we had to integrate our firm-level data from the two statistical sources described above with

additional information from the 44-industry input–output table of 1992 for the Italian

120

115

110

105

100

95

90

85

80

1984

1985

1986

1987

1988

1989

1980

1981

1982

1983

1984

1985

1986

1987

1988

Export Exchange Rate

ImportExchangeRate

FIGURE 1Real Import and Export Exchange Rates (1998 = 100)

Source: Bank of Italy.



economy. In particular, precisely as in Nucci and Pozzolo (2001, 2010), we extracted for each

industry j the value of purchases of imported intermediate inputs, IMj1992, and the value of

purchases of total intermediate inputs, TEj1992. Then, using time series information on import

demand and production for each industry, j, we updated backward and forward from the

input–output table that refers to one year only both the value of imported input purchases and

the value of total input purchases (domestically produced and imported). Last, we combined

these industry-level information with our data at the firm level on expenditure for intermediate

inputs, TEit. In particular, we computed the share of costs on imported inputs in total input

expenditure, ait, as follows: ait = [(IMjt/TEjt)TEit]/[TEit + LCit], where IMjt is the value of

intermediate inputs imported by industry j, which denotes the industry to which the generic

firm i belongs, and TEjt is the value of total expenditure for intermediate inputs of industry j(to which firm i belongs). TEit is the firm-level and time-varying value of total expenditure

for intermediate inputs and LCit is firm i’s expenditure for labour services (see Campa and

Goldberg, 1997, for a similar approach). Thus, while we have data on the demand for inputs

that do vary across firms in the same industry, we admittedly have to supplement this infor-

mation with industry-level information and, as a result, the share ait incorporates elements

which are not entirely firm-specific. In order to provide some indication of the variability of

ait between firms in the same industry, we have computed the standard deviation of aitbetween firms in the same industry in each given year. By comparing these figures on the dis-

persion of ait across firms of the same industry, the picture that emerges is that the industry

standard deviations of the share ait range, in every year of the sample, from a minimum value

of 0.01 to a maximum value of 0.03. Due to the data constraint in measuring the degree of

firm’s reliance on imported inputs, ait, we are aware that our measure does not accurately

reflect the within-industry heterogeneity of firms’ import decisions and this suggests caution

in the interpretation of the results. Arguably, however, this imperfect measurement is likely to

result in an attenuation bias in the parameter estimate, thus reinforcing the significance of our

results.

Following the approach developed by Domowitz et al. (1986), we computed a time-varying

measure of each firm’s market power as the ratio of the firm’s value added net of labour com-

pensation to the value of firm’s total production. Ideally, we would use distinct destination-

specific mark-up ratios, for example in the home and the foreign markets. Since our data do

not allow us to derive them, we construct a firm’s average measure of market power for each

year.

The exchange rates used in the empirical analysis are the permanent components of the

export and import real effective exchange rates of the Italian lira constructed by considering

24 different bilateral exchange rates, based on producer price indexes (see Banca d’Italia,

1998). The permanent component has been derived applying the Beveridge and Nelson (1981)

procedure that decomposes a non-stationary series into its permanent and transitory compo-

nents (see Nucci and Pozzolo, 2010, for details). Figure 1 shows the time profile of the

monthly data on import and export real exchange rates in the period analysed (with an

increase in the exchange rates amounting to a real appreciation). While they exhibit a very

similar pattern, some differences emerge in the mid-1980s and at the beginning of the 1990s.

Table 1 documents some descriptive statistics for the variables used in our empirical analy-

sis. During the sample period, the average rate of annual changes of the permanent component

of the import exchange rate is �0.29 per cent. However, the standard deviation is 5.40, sug-

gesting a non-negligible variability, consistent with the evolution of the exchange rate of the

lira described above. In the case of the export exchange rate, the mean rate of change is



�0.09 per cent, but the value of the standard deviation points to an even larger variability.

The average degree of a firm’s dependence on imported inputs is 13.9 per cent, and the aver-

age share of export revenues in total revenues is more than double that of import dependence,

at 29.8 per cent. Both distributions are skewed to the right, consistent with the findings of the

empirical literature also for other countries. Interestingly, even at the 25th percentile of the

distribution, the share of export revenues is non-negligible (4.9 per cent). The standard devia-

tion of import and export shares is also sizeable, amounting, respectively, to 5.6 and 26.9

per cent. Most of the sample variability reflects differences across firms, as shown by the the

standard deviation of each firm’s import and export shares during the sample period, respec-

tively 1.5 and 6.1 per cent. The mean and median mark-ups are around 9 per cent, and they

also show substantial dispersion. The average rate of growth of wages per employee is around

1.2 per cent, that of wages per hour about 0.4. Their standard deviations across firms and

years are both very high, between 15 and 20 per cent. Labour productivity per employee and

per hour increased at an average rate of 1.7 and 0.9 per cent per year, respectively, with a

standard deviation in both cases of about 25 per cent. Finally, the share of blue collars is

about 70 per cent, with a standard deviation of 18 per cent.

b. The Econometric Specification

To assess empirically the wage response to exchange rate fluctuations, we rely on a base-

line equation which has the following specification:

TABLE 1Summary Statistics

Variable Mean SD Median 25thpercentile

75thpercentile

Import exchange rate (growth) �0.29 5.40 0.17 �2.29 3.07Export exchange rate (growth) �0.09 5.75 �0.25 �2.72 4.36Imported input share (a) 13.90 5.62 12.78 10.23 16.40Export share (v) 29.77 26.93 23.65 4.89 48.90Firm-level variability of imported input share 1.45 0.91 1.31 0.74 2.01Firm-level variability of export share 6.12 6.11 4.45 1.84 8.41Firm-level variability of import-side interaction 0.74 0.45 0.71 0.37 0.97Firm-level variability of export-side interaction 1.65 1.60 1.19 0.31 2.63Mark-up 8.91 25.99 9.34 5.60 13.83Wage per employee (growth) 1.18 16.96 0.90 �2.96 4.82Wage per hour (growth) 0.37 18.61 0.53 �3.71 4.78Total sales (growth) 0.08 23.64 0.34 �8.58 9.30Labour productivity per employee (growth) 1.72 25.45 1.39 �7.69 10.37Labour productivity per hour (growth) 0.91 25.48 0.80 �8.08 9.64Blue collars (share) 69.83 18.30 74.25 61.49 82.46

Notes:(i) All variables are expressed as percentages.(ii) Firm-level variability of imported input share and export revenues is calculated from the time series standarddeviation of each firm’s imported input share and export share.(iii) Firm-level variability of the interactions is calculated from the time series standard deviations of the interactionbetween each share and its respective exchange rate variation.(iv) The sample period considered is 1984–98.



Dwit ¼ b0 þ b1vit�1Dpeert þ b2ait�1Dpmert þ b3vit�1 þ b4ait�1þþ b5Dsit�1 þ b6MKUPit�1 þ b7Dnit�1 þ b8Dlpit�1 þ b0Zit þ ki þ uit; (1)

where lower case letters denote the logarithmic transformation of the variable. Wit is the aver-

age wage in real terms paid by firm i at time t (we alternatively use the wage per employee

and the wage per hour); vit�1 is the share of sales in foreign markets over total sales at time

t�1; ait�1 is the share of production costs for imported inputs in total costs. Both these shares

enter the specifications with a one-period lag in order to mitigate the simultaneity bias possi-

bly originating from the impact of exchange rate swings on the extent of firms’ external expo-

sure. PEERt and PMERt are the permanent components of, respectively, export and import

real effective exchange rates, defined so that an increase in the exchange rate is an apprecia-

tion; MKUPit is an index of firms’ market power; Nit is the amount of labour input (according

to the definition of the real wage adopted, we alternatively measure it as the number of

employees and the number of total hours); Sit is the value of real sales; LPit is the level of

labour productivity, computed as gross output at constant prices over labour input (alterna-

tively measured as the number of employees and the number of total hours); Zit is a vector of

dummy variables controlling for the different years, industries, sizes and geographical loca-

tions in which each firm operates. Finally, the specification controls for individual firm latent

heterogeneity including firm-level fixed effects, ki; uit are the disturbance terms, for which we

assume that E(uit) = E(uituis) = 0, for all t 6¼ s. Precisely as in Campa and Goldberg (1999,

2001) and Nucci and Pozzolo (2001, 2010), the empirical specification is in first-differences

in the light of the non-stationarity of the exchange rate time series.

The dynamic equation (1) is a reduced form, but it allows us to directly test the theo-

retical predictions discussed earlier. The key variables for characterising the wage adjust-

ment in response to currency swings are (i) vit�1�Dpeert, the interaction term of the export

share of sales lagged by one period with the export exchange rate variation and (ii)

ait�1�Dpmert, the interaction term of the lagged share of expenditure for imported inputs in

total purchases with the import exchange rate variation. The advantage of this approach

using interaction terms is that it allows the estimated sensitivity of wages to currency

movements to vary across firms and over time, depending on the evolving external orienta-

tion of each specific firm on both the revenue and cost side (see Campa and Goldberg,

2001; Nucci and Pozzolo, 2010). In other words, the empirical framework of equation (1)

allows us to estimate a time-varying and firm-specific response of firms’ labour compensa-

tion to exchange rate oscillations, distinguishing between the export revenues and the

imported input cost channels. Of course, the export share of sales and the share of

imported inputs are also inserted in the specification as single regressors in isolation.

The lagged value of changes in the firms’ real sales is included to control for demand condi-

tions, and the lagged value of the mark-up ratio to control for any effect of marginal profitability

on wages that is independent of exchange rate oscillations. Consistent with the theoretical set-up

of Nucci and Pozzolo (2010), the specification includes the lagged value of change in employ-

ment (or hours) to control for the adjustment lags that typically characterises the labour market.

As in Nucci and Pozzolo (2001, 2010), we use the generalised method of moments (GMM)

estimator for panel data models, which was shown to be efficient within the class of instrumental

variable estimators (Arellano and Bond, 1991). Indeed, since in equation (1), the lagged values

of the mark-up and of change in employment (hours) and sales are likely to be correlated with

the firm-specific fixed effects, ki, the endogeneity of these regressors might cause inconsistency

of the parameters estimated with standard panel methods. We therefore treat these variables as



endogenous and use the GMM estimator to ensure consistency of parameter estimates. Specifi-

cally, following Arellano and Bover (1995) and Blundell and Bond (1998), we rely on the sys-

tem GMM panel estimator, which augments the Arellano and Bond (1991) methodology by

building a system of two equations: the original equation (i.e. equation (1)) and the transformed

equation where a suitable transformation of the data allows to expunge fixed effects.13 Compared

with the Arellano-Bond estimator, system GMM allows us to exploit an even larger number of

moment conditions, by resorting to a larger instrument set and thereby increasing efficiency. In

particular, building on Arellano and Bover (1995), Blundell and Bond (1998) develop a system

estimator where the transformed equation is instrumented with original, untransformed variables

(precisely as in the Arellano-Bond method), while the original equation is instrumented with

transformed variables. In the estimation, we utilise as GMM type of instruments for the endoge-

nous variables the lagged values of changes in real sales and employment (or hours) and the

lagged values of the mark-up. The validity of our specification and of the set of instruments is

ascertained by conducting (i) the Hansen test of over-identifying restrictions, which seeks to ver-

ify the orthogonality between instrumental variables and the disturbance terms and (ii) the Arel-

lano-Bond test for second-order serial correlation of residuals of the transformed equation. We

now turn to present and discuss the estimation results.

4. EMPIRICAL RESULTS

a. The Baseline Specification

The results from estimating equation (1) are presented in Table 2. The effect of exchange

rate fluctuations on labour compensation in each firm is statistically significant in terms of

both the average wage per employee (column 1) and per hour (column 2), and the estimated

coefficients of the two interaction terms, capturing the effect of currency swings on wages

through, respectively, the foreign sale channel and the imported intermediate inputs channel

have the expected sign. If we focus on the response of the (average) wage per employee

(column 1), the estimated coefficient of vit�1�Dpeert is �0.173 with a standard error of 0.048,

and the estimated coefficient of ait�1�Dpmert is 0.730 with a standard error of 0.232. In the

case of wage per hour (column 2), the estimated coefficient of vit�1�Dpeert is �0.689 with a

standard error of 0.124, and the estimated coefficient of ait�1�Dpmert is 0.614 with a standard

error of 0.257. These results support the theoretical prediction that a currency depreciation,

that is, a negative variation over time of both peert and pmert causes a rise in average real

wages through the foreign sales side of the balance sheet and to a decline through the expen-

diture side. Both effects are estimated to be stronger, respectively, the higher is the firm’s

international exposure through exports, that is, the higher is vit�1, and the higher is the firm’s

reliance on imported inputs, that is, the higher is ait�1.14

A natural question that arises from our evidence is whether an exchange rate appreciation

leads to a rise or fall of the labour compensation set by the firm. As argued in Nucci and

13 We conduct estimation using the xtabond2 program for Stata developed by Roodman (2009).14 In our empirical specification, the effect of the average exchange rate is captured by the time dum-mies. However, we have also verified that the results are confirmed when the time dummies areexcluded from the specifications and are replaced by the import or export real exchange rate or, sincethe two are strongly correlated, by their first principal component calculated on the sample period. Thefindings from these regressions are not reported here but are available upon request.



TABLE 2The Effect of Exchange Rate Change on Wages

Variable (1) (2)Labour compensationper employee: Dwit

Labour compensationper hour: Dwhit

ait�1�Dpmert 0.730*** 0.614**(0.232) (0.257)

vit�1�Dpeert �0.173*** �0.689***(0.048) (0.124)

ait�1 0.288*** 0.178***(0.034) (0.038)

vit�1 �0.002 �0.003(0.003) (0.003)

MKUPit�1 0.101*** 0.009(0.011) (0.012)

Dsit�1 �0.010 �0.076***(0.002) (0.004)

Dnit�1 0.038*** 0.280***(0.008) (0.009)

Dlpit�1 0.022*** 0.102***(0.003) (0.004)

Year dummies 1081.1(0.00)

360.9(0.00)

Industry dummies 113.8(0.00)

31.0(0.00)

Geography dummies 24.1(0.00)

8.32(0.08)

Firm size dummies 1.74(0.78)

34.23(0.00)

Hansen test of over-identifying restrictions: 369.7(0.14)

394.2(0.10)

Test for second-order serial correlation 1.78(0.08)

�0.74(0.46)

Number of observations 6,579 6,220

Notes:(i) The system GMM dynamic panel methodology is used for estimation.(ii) ait�1 is the share of expenditure for imported inputs and vit�1 is the export share of sales.(iii) Dpmert and Dpeert are the (log) changes in the permanent component of respectively, the import and exportexchange rate.(iv) Dsit�1 is the (log) variation of real sales and MKUPit�1 is the firm’s mark-up ratio.(v) Dnit�1 is the (log) variation of labour input and is measured as number of employees in column 1 and as numberof hours in column 2.(vi) Dlpit�1 is the (log) variation of labour productivity.(vii) Size dummies refer to 50–99, 100–199, 200–499, 500–999, ≥1,000 employees.(viii) Geographical dummies refer to north-west, north-east, centre, south, islands.(ix) For each group of dummies, we report the value of Wald test of their joint significance and the associatedp-value.(x) Standard errors are corrected for heteroscedasticity and reported in parentheses.(xi) The instrument set includes lagged values of changes of labour inputs, sales and the mark-up dated t�2 and earlier.(xii) Hansen is a test of over-identifying restrictions asymptotically distributed as a v2.(xiii) Sample period: 1984–1998.(xiv) *** denotes significance at the 1% confidence level, ** at the 5% confidence level and * at the 10% confidence level.



Pozzolo (2010), our empirical framework is not the most suitable one for ascertaining the

aggregate effect of exchange rate swings on wages. On the contrary, the primary advantage of

our approach is precisely that it allows to capture the firm-specific relevance of each transmis-

sion channel. This implies that a firm-specific wage response to exchange rates can be esti-

mated on our microeconomic data for each period. To do so, consider first the mean value of

the regression sample of both the export share of sales, vit, and the share of imported input

costs, ait, equal to 0.298 and 0.139, respectively. If we evaluate the shares reflecting external

orientation at these mean values, and use, for example, the estimation results reported in col-

umn 2 of Table 2, the estimated elasticity of wage per hour to exchange rate change is

�0.12. This means that the effect of a 1 per cent currency depreciation on the hourly wage

for a hypothetical firm with this type of foreign exposure is a 0.12 per cent real wage rise.

Similarly, measuring foreign exposure at median values, the effect of a 1 per cent currency

depreciation is a 0.08 per cent real wage rise.

However, if instead of considering a hypothetical firm exhibiting average shares, we con-

sider firms’ heterogeneity in terms of external orientation, by focusing on firms at the median

value of the distribution of export (import) shares and, respectively, at the 10th and 90th per-

centiles of the distribution of import (export) shares, we obtain rather different results. For the

firm at the 10th percentile of the distribution of the share of imports (with a = 0.08), a 1 per

cent currency depreciation determines a 0.11 per cent increase in real wages per hour, while

for the firm at the 90th percentile (with a = 0.22), it determines a wage decrease of 0.03 per

cent. For the firm at the 10th percentile of the distribution of the share of exports (with vbelow 0.01), a 1 per cent currency depreciation determines a 0.08 per cent drop of real wages

per hour, but for the firm at the 90th percentile (with v = 0.71), it determines a rise of the real

wages per hour of 0.42 per cent. Figure 2 shows how the estimated effect on wages of a 1

per cent currency depreciation varies with the export share of sales and with the share of costs

1.00

0.00

–1.000.00

0.1 0.2 0.3 0.40.5

0.60.7

0.80.9

0.20

0.400.60

0.801.00

1

Export Share of SalesImported Inputs Shareof Total Costs

–1.00–0.00 0.00–1.00

Percentage Changeof Real Wages

FIGURE 2The Heterogeneous Impact of a 1 Per Cent Depreciation on Wages Depending

on the External Orientation of Firms



of imported inputs on total costs. These results point to a high degree of heterogeneity across

firms, suggesting that the analysis of average values may hide substantial recomposition

effects within and across industries.

The estimation results also document that the firm’s profit margin, as measured by the

mark-up ratio, MKUPit�1, exerts a positive effects on wages, and so does the lagged change

in employment (or hours), Dnit�1 and in labour productivity, Dlpit�1. On the contrary, the

change in total sales, Dsit�1, which was also included in the specification as a control variable,

has a negative and statistically significant effect on both the wage per employee and per hour.

As already discussed above, the specification includes a number of dummy variables. We

report the value of the Wald tests for the joint significance of each group of dummies, indicat-

ing that all these effects are statistically significant. Evidence on the validity of our baseline

specification in both columns 1 and 2 is provided by the values of the Hansen statistic for

over-identifying restrictions and of the test for the absence of second-order serial correlation

of residuals.15

b. The Role of Market Power

As discussed in Section 2, a notable prediction of the theoretical framework proposed in

the literature is that, for a given international exposure, the sensitivity of wages to exchange

rate fluctuations is larger for firms with a lower market power in the product markets. We ver-

ify empirically this hypothesis estimating our baseline specification on two different subsam-

ples obtained using the median value of firms’ mark-ups as the separating threshold (see

Campa and Goldberg, 2001; Nucci and Pozzolo, 2010). The estimation results, reported in

Table 3, show that the effect of exchange rate swings on (average) labour compensation both

per employee and per hour is stronger for firms with a lower mark-up ratio. If we focus, for

example, on the impact on compensation per hour, the estimated effect through the cost side

is 1.773 with a standard error of 0.360 for the firms with relatively low pricing power, which

is indeed larger compared with the corresponding one for firms with higher pricing power (in

this case, it is 0.700 with a standard error of 0.292; see columns 3 and 4 of Table 3). By the

same token, the estimated impact of exchange rate through the revenue side is �0.293 with a

15 In unreported regressions, available upon request, we have extended our analysis to study how twocharacteristics of the workforce impact on our results. These are the incidence of part-time workers andthe skill intensity of workers, as proxied by the share of white-collar employees in each firm. Our data,unfortunately, do not include direct information on the share of part-time workers in each firm. We havetherefore approximated their incidence with the number of hours per employee in every year. Our empir-ical framework was then modified by replacing the interaction terms between exchange rate variationsand the variables describing the firm’s international exposure with new interactions terms that comprisealso the firm’s amount of hours per employee, in one case, and the firm’s share of white-collar workers,in the other case. Of course, we have also included the value of each interacted term in isolation.Although our theoretical framework does not provide an exact prediction of the impact of these charac-teristics on the relationship between exchange rates and wages, we argue that the wages of part-timeworkers and of skilled workers are likely to be more flexible and therefore more sensitive to exchangerate swings. Indeed, the firm’s reliance on performance-pay schemes has increased since the mid-1980scausing a higher flexibility of wages and this is especially true among executives and white-collarsworkers (Lemieux et al., 2009). On the other hand, for part-time workers, the contractual arrangementsare likely to be shorter in length, making their wages more exposed to shocks. The regression resultslend support to these hypotheses as we estimate a positive coefficient of the new interaction terms onthe imported input side and a negative coefficient of the new interaction terms on the export side.



standard error of 0.044 in the case of firms with lower market power, while it is �0.227 (with

a standard error of 0.136) for firms exhibiting higher pricing power in the product markets.

Similar results are obtained considering the impact on wages per employee.

TABLE 3Exchange Rate, Market Power and Wage Adjustment (I)

Variable (1) (2) (3) (4)Compensation per Employee: Dwit

Degree of Market PowerCompensation per Hour: DwhitDegree of Market Power

Low High Low High

ait�1�Dpmert 2.937*** 0.740*** 1.773*** 0.700**(0.668) (0.269) (0.360) (0.292)

vit�1�Dpeert �0.393*** �0.113* �0.293*** �0.227*(0.067) (0.065) (0.044) (0.136)

ait�1 0.041*** 0.128*** 0.508*** 0.172***(0.004) (0.041) (0.047) (0.049)

vit�1 �0.007* 0.004 �0.020* 0.005(0.004) (0.004) (0.004) (0.004)

Dsit�1 �0.006*** �0.035*** �0.022*** �0.024***(0.002) (0.003) (0.001) (0.002)

MKUPit�1 0.001 0.128*** 0.003 0.150***(0.010) (0.026) (0.008) (0.024)

Dnit�1 0.048** �0.014 0.128*** 0.093***(0.009) (0.013) (0.005) (0.009)

Dlpit�1 0.041** 0.041*** 0.042*** 0.023***(0.004) (0.005) (0.003) (0.005)


598.7(0.00)

581.8(0.00)

297.9(0.00)


49.4(0.00)

136.2(0.00)

29.5(0.01)


22.1(0.00)

22.3(0.00)

34.2(0.00)


14.4(0.01)

43.8(0.00)

52.1(0.00)

Hansen test ofOver-identifyingrestrictions

297.4(0.16)

267.3(0.06)

386.0(0.23)

291.00(0.33)

Test for second-orderserial correlation 0.72

(0.47)1.98(0.05)

�1.08(0.28)

�0.17(0.87)

Number ofobservations

3,334 3,245 3,122 3,098

Notes:(i) See Table 1.(ii) The system GMM dynamic panel methodology is used for estimation.(iii) The sample is split based on the degree of firms’ market power.(iv) The threshold criterion is the median of firms’ mark-up.(v) Variables in lower case letters denote their logarithmic transformation.(vi) Sample period: 1984–98.(vii) *** denotes significance at the 1% confidence level, ** at the 5% confidence level and * at the 10% confidencelevel.



To analyse this issue in more detail, we also modify the baseline specification. In particu-

lar, we replace the two key explanatory variables in our regression, namely the interaction

terms between exchange rate variations and the variables of firm’s international exposure

(respectively, vit�1�Dpeert and ait�1�Dpmert) with two new interaction terms that comprise

also the firm’s level of market power, that is, MKUPit�1, as a further multiplicative term.

Specifically, the new regressors are (i) vit�1�Dpeert�(1�MKUPit�1) and (ii) ait�1�Dpmert�(1�MKUPit�1).

TABLE 4Exchange rate, Market Power and Wage Adjustment (II)

Variable Real Wages: Dwit

(1) (2)Labour Compensationper Employee: Dwit

Labour Compensationper Hour: Dwhit

ait�1�Dpmert�(1�MKUPit�1) 0.800** 1.191**(0.343) (0.605)

vit�1�Dpeert�(1�MKUPit�1) �0.259*** �0.548***(0.072) (0.126)

ait�1 0.258*** 0.241***(0.042) (0.036)

vit�1 �0.003 �0.0004(0.004) (0.003)

Dsit�1 �0.013*** �0.047***(0.003) (0.002)

MKUPit�1 0.137*** 0.034***(0.017) (0.012)

Dnit�1 0.046*** 0.190***(0.011) (0.006)

Dlpit�1 0.028*** 0.079***(0.005) (0.004)


462.4(0.00)


48.8(0.00)


19.8(0.00)


51.9(0.00)

Hansen test of over-identifying restrictions: 244.0(0.13)

413.9(0.09)

Test for second-order serial correlation 1.83(0.07)

�0.55(0.58)

Number of observations 6,579 6,220

Notes:(i) See Table 1.(ii) The system GMM dynamic panel methodology is used for estimation.(iii) Variables in lower case letters denote their logarithmic transformation.(iv) Sample period: 1984–98.(v) *** denotes significance at the 1% confidence level, ** at the 5% confidence level and * at the 10% confidencelevel.



The theoretical prediction is that a depreciation (i.e. a negative value of Dpeert) would

increase wages through the revenue side, and the impact is expected to be stronger the higher

the share of exports from sale and the lower the price mark-up. Therefore, a negative estimate

of the parameter for the interaction term for the revenue side would lend empirical support to

this prediction. Indeed, Table 4 shows that the estimated coefficient is �0.259 (with a stan-

dard error of 0.072) when we consider labour compensation per employee and �0.548 (with a

standard error of 0.126) when we consider labour compensation per hour. On the other hand,

we expect a depreciation to reduce wages through the cost side with the size of the impact

being higher (in absolute value) the higher is the share of expenditure on imported input over

total costs and the lower is the price mark-up. Estimating a positive parameter for the interac-

tion term on the cost side would support the model’s prediction, and this is the result we

obtain. Indeed, the estimated coefficient is positive and statistically significant when we con-

sider both labour compensation per employee (0.800 with a standard error of 0.343; see Table

4, column 1) and labour compensation per hour (1.191 with a standard error of 0.605; see

Table 4, column 2).

c. The Different Degree of Import Penetration

To further characterise the linkage between exchange rate and wages, we analyse some

additional features that may contribute to shape the sensitivity of firm-level wages to currency

shocks.

So far, we have emphasised foreign exposure of a firm as captured by both the extent of

the exporting activity and the incidence of imported intermediate inputs. However, the firm is

exposed to international competition also in the domestic product markets, and the degree to

which this happens depends on (i) the extent of import penetration in the domestic industry to

which a firm belongs and (ii) the share of firm’s sales in the domestic market over total sales.

As we discussed in Section 2, the more relevant are import penetration and the firm’s

exposure to domestic revenues, the more severe is the competitive pressure exerted by foreign

producers in the firm’s domestic market. This channel introduces another degree of difference

across firms in the estimated impact of exchange rate movements on labour compensation.

To account for this effect, following Nucci and Pozzolo (2010), we therefore consider the

following specification:

Dwit ¼ b0 þ b1vit�1Dpeert þ b2ait�1Dpmert þ b3vit�1 þ b4ait�1 þþb5Dsit�1

þ b6MKUPit�1 þ b7Dnit�1 þ b8Dlpit�1 þXK

j¼1

cjð1� vit�1ÞIPjt�1DpeertDj

� �

þ b0Zit þ ki þ uit; (2)

which is identical to equation (1), except for the inclusion of the summation term, that is

composed by: (i) Dj, which is a dummy variable for each industry j (with j = 1,2,. . .K)and is equal to one if firm i belongs to industry j and zero otherwise; (ii) IPjt�1, the

(lagged value of the) industry j’s import penetration ratio, as measured by the share of

imports of products j over domestic demand for those products, obtained as the industry’s

sales plus the imports of products of industry j minus the industry’s exports; and (iii)

(1�vit�1), that is the ratio of domestic sales to total sales and reflects the degree of firm’s

exposure on the domestic product market. All these variables interact with the export



exchange rate change, so that for each industry, we can estimate an industry-specific

coefficient, cj, associated with the corresponding interaction term. These coefficients are

expected to be negative, because a currency depreciation (a fall of peert) augments the

firm’s competitiveness in the domestic market, thus increasing its profitability through the

domestic revenue side and thereby the equilibrium wages. Moreover, the size of this effect

will vary across firms depending on the relevance of domestic sales in total sales and on

the extent of import penetration. This hypothesis is confirmed by the results from

estimating the model in equation (3) that are reported in Table 5. The coefficients of the

two key interaction terms have the expected sign and are statistically significant. In

TABLE 5Exchange Rate, Import Penetration and Wage Adjustment

Variable Labour CompensationPer Employee: Dwit

ait�1�Dpmert 4.120**(1.949)

vit�1�Dpeert �3.577***(0.950)

(1�vit�1)�Dpeert�IP1t�1�D1, (1�vit�1)�Dpeert�IP2t�1�D2,.......,(1�vit�1)�Dpeert�IPKt�1�DK

Wald test74.7(p�val:0.00)

ait�1 0.221***(0.056)

vit�1 �0.010***(0.004)

Dsit�1 �0.002(0.006)

MKUPit�1 0.058***(0.014)

Dnit�1 0.012(0.017)

Dlpit�1 �0.016(0.011)

Year dummies 321.6 (0.00)Industry dummies 104.7 (0.00)Geography dummies 8.1(0.09)Firm size dummies 3.2(0.36)

Hansen test of over-identifying restrictions: 219.6 (0.73)Test for second-order serial correlation 1.98 (0.05)Number of observations 5,184

Notes:(i) See Table 1.(ii) The system GMM dynamic panel methodology is used for estimation.(iii) IPjt is the value of import penetration experienced by industry j (to which firm i belongs) in the year t.(iv) Dj is the j-th industry dummy, taking the value of one if firm i belongs to industry j and zero otherwise.(v) The Wald statistic associated with the variables (1 � vit � 1)�Dpeert�IPjt�1�Dj (j = 1,2,..., K) tests for the jointhypothesis that their coefficients are equal.(vi) Variables in lower case letters denote their logarithmic transformation.(vii) Sample period: 1984–98.(viii) *** denotes significance at the 1% confidence level, ** at the 5% confidence level and * at the 10% confidencelevel.



analysing the industry-specific values of the estimated coefficients cj, capturing the effect

of exchange rate on wages through import penetration, we first test whether they are

different among each others, conducting a Wald test. Indeed, the null hypothesis of

identical industry-specific coefficients is strongly rejected (with a p-value of 0.00). In

Table 6, we also report the values for each industry j of the estimated wage sensitivity to

exchange rate through the import penetration channel, which we computed combining the

value of the coefficient cj with (i) the corresponding industry-specific average value across

firms and over time of (1�vit�1) and (iii) the time average of IPjt for the industry. In

each of the 15 industries, the estimated wage elasticity to exchange rate through the

import penetration channel has the expected negative sign. To gauge whether these

estimates are sensible, in Table 6 we compare these estimates of the elasticities with the

corresponding values of import penetration for each industry, together with the associated

rank. As in Nucci and Pozzolo (2010) for the employment elasticity, the estimated impact

of exchange rate on wages through the domestic sales side is larger for industries exhibit-

ing a higher degree of import penetration. The strongest effects are recorded for electrical

machinery and for computers and office equipments. For the latter industry, the extent of

import penetration is very high (58 per cent; ranked no. 1), and for electrical machinery,

it is also sizeable (29 per cent; ranked no. 2). We have also computed the Spearman’s

rank correlation between the (absolute values of the) estimated industry-specific wage

responses on the domestic sale side and the indexes of import penetration of the

TABLE 6Import Penetration across Industries and the Sensitivity of Wage to Exchange Rate

Industry ImportPenetration

Wage ResponseThrough ImportPenetration

Value Rank Estimate Rank

Transformation of non-metalliferous minerals 0.10 14 �0.46 15Chemicals 0.35 3 �2.66 4Metals 0.05 15 �2.28 7Machinery for industry and agriculture 0.25 5 �1.30 10Computers, office equipments, precision instruments 0.58 1 �3.28 1Electrical machinery 0.29 4 �3.10 2Motor cars and other transport equipments 0.54 2 �2.73 3Food and tobacco products 0.16 7 �2.46 6Textiles 0.16 8 �1.78 8Leather and footwear 0.17 6 �2.52 5Clothing 0.14 12 �1.27 11Wood and furniture 0.15 11 �1.74 9Paper and publishing 0.12 13 �1.14 12Rubber and plastic products 0.15 9 �0.59 14Other manufactures 0.15 10 �0.84 13

Notes:(i) To derive import penetration for each industry and the industry-specific estimated response of wages to exchangerate through import penetration see the discussion in the text.(ii) The industry-specific wage responses are obtained from the estimation results documented in Table 4; the ranksof these estimated effects pertains to their absolute values.



corresponding industries. With a value of the Spearman correlation of 0.75, the null

hypothesis that these two variables are independent is rejected at the 1 per cent level of

statistical significance.16

5. CONCLUDING REMARKS

Using data on a representative panel of Italian manufacturing firms, we find that exchange

rate fluctuations do affect the labour compensation set by each firm. Similar to the analysis in

Nucci and Pozzolo (2010) on the response of employment and hours conducted on the same

microeconomic data, we find that the direction and size of wage adjustment is determined by

the external orientation of each firm on both the revenue and cost side of its balance sheet.

Through the revenue side, a currency depreciation affects firm’s profitability and thereby the

equilibrium outcome of wages. The latter are shown to rise along this channel, and the effect

is estimated to be larger the higher is the firm’s exposure to revenues from exporting. On the

other hand, a depreciation leads to a reduction in the firm’s wages through the channel of

expenditure for imported inputs, and the effect is larger the higher is the firm’s reliance on

imported vis-�a-vis domestically produced inputs. Our results indicate that, for a given type of

firm’s international exposure, the responsiveness of wages to exchange rate is more pro-

nounced for firms with a lower degree of pricing power.

To provide further characterisations of the wage sensitivity, we also document that other

transmission channels introduce a significant source of heterogeneity across firms in the

response of wages to exchange rate swings. These include the extent of competition in

the domestic marked exerted by foreign producers, as measured by the import penetration in

the domestic industry to which a firm belongs.

Although our empirical findings pertain to manufacturing firms, and therefore extending

the scope of their validity to the whole economy would be questionable, our results highlight

a further channel through which currency oscillations can have significant effects on the real

economy. Most important, these effects show wide differences, even in their sign, depending

on individual firm’s characteristics. Beyond average effects, currency oscillations can there-

fore have substantial redistributive effects within and across industries.

16 We have also focused in more details on the effects through the cost side, with a possible source ofspecificity being the different degree of substitutability between imported and domestically producedintermediate inputs. After a depreciation, that increases the price of imported inputs in the domestic cur-rency, a firm may decide to replace its imports of intermediate goods with similar goods that are domes-tically produced. Arguably, the extent to which this happens may reflect technological and organisationalcharacteristics of the firm that are shared by the firms in the same industry. If the degree of this substi-tutability is relatively high (low), then the wage sensitivity to exchange rates through the cost side chan-nel would be relatively low (high). To investigate this issue, we estimate an equation that is identical toequation (1), except for the effect of ait�1DPMERt on wages that is estimated separately for each indus-try. To ascertain whether the differences between the estimated coefficients summarising the cost-sideeffects of exchange rate on wages are significant, we performed a Wald test for the null hypothesis thatthese differences are equal to zero. The value of the test is 38.5 with a p-value of 0.00, indicating thatdifferences across industries in the wage sensitivity through the cost side are statistically significant.However, our baseline conclusion that the average effect of an exchange rate depreciation through theimport side is that of determining a fall of real wages is also in this case confirmed.



REFERENCES

Abowd, J. A. and T. Lemieux (1993), ‘The Effects of Product Market Competition on Collective Bar-gaining Agreements: The Case of Foreign Competition in Canada’, Quarterly Journal of Economics,108, 4, 983–1014.

Arai, M. (2003), ‘Wages, Profits, and Capital Intensity: Evidence from Matched Worker-firm Data’,Journal of Labor Economics, 21, 3, 593–618.

Arellano, M. and S. R. Bond (1991), ‘Some Tests of Specification for Panel Data: Monte Carlo Evidenceand an Application to Employment Equations’, Review of Economic Studies, 58, 2, 277–97.

Arellano, M. and O. Bover (1995), ‘Another Look at the Instrumental Variable Estimation of Error-components Models’, Journal of Econometrics, 68, 1, 29–51.

Banca d’Italia (1998), ‘Nuovi Indicatori di Tasso di Cambio Effettivo Nominale e Reale’, BollettinoEconomico, 30, 1–8.

Banca d’Italia (2009), ‘Rapporto Sulle Tendenze Nel Sistema Produttivo Italiano’, Questioni di Econo-mia e Finanza Occasional Papers, (Roma: Banca d’Italia).

Berman, E., J. Bound and Z. Griliches (1994), ‘Changes in the Demand for Skilled Labor Within USManufacturing: Evidence from the Annual Survey of Manufacturers’, Quarterly Journal of Econ-omics, 109, 2, 367–97.

Bernard, A. B., J. B. Jensen, S. J. Redding and P. K. Schott (2007), ‘Firms in International Trade’,Journal of Economic Perspectives, 21, 1, 105–30.

Beveridge, S. and C. R. Nelson (1981), ‘New Approach to Decomposition of Economic Time Series intoPermanent and Transitory Components with Particular Attention to Measurement of the ‘BusinessCycle’’, Journal of Monetary Economics, 7, 2, 151–74.

Blanchflower, D. G., A. J. Oswald and Sanfey, P. (1996), ‘Wages, Profits, and Rent-sharing’, QuarterlyJournal of Economics, 111, 1, 227–52.

Blundell, R. and S. Bond (1998), ‘Initial Conditions and Moment Restrictions in Dynamic Panel DataModels’, Journal of Econometrics, 87, 1, 115–43.

Campa, J. and L. S. Goldberg (1997), ‘The Evolving External Orientation of Manufacturing: a Profile ofFour Countries’, Economic Policy Review Federal Reserve bank of New York 3, 1, 53–81.

Campa, J. and L. Goldberg (1999), ‘Investment, Pass-through and Exchange Rates: A Cross-countryComparison’, International Economic Review, 40, 2, 287–314.

Campa, J. and L. Goldberg (2001), ‘Employment versus Wage Adjustment and the US Dollar’, Reviewof Economics and Statistics, 83, 477–89.

Domowitz, I., R. G. Hubbard and C. Petersen (1986), ‘Business Cycles and the Relationship betweenConcentration and Profit Margins’, Rand Journal of Economics, 17, 1–17.

Dornbusch, R. (1987), ‘Exchange Rates and Prices’, American Economic Review, 77, 93–106.Goldberg, L. and J. Tracy (2000), ‘Exchange Rates and Local Labor Markets’, in R. Feenstra (ed.), The

Impact of International Trade on Wages (Chicago, IL: University of Chicago Press), 269–308.Goldberg, L. and J. Tracy (2001), ‘Exchange Rates and Wages’, Staff Report 116 (New York: Federal

Reserve Bank of New York).Greenaway, D., R. Kneller and X. Zhang (2010), ‘The Effect of Exchange Rates on Firm Exports: The

Role of Imported Intermediate Inputs’, The World Economy, 33, 8, 961–86.Hildreth, A. K. G. and A. J. Oswald (1997), ‘Rent-sharing and Wages: Evidence from Company and

Establishment Panels’, Journal of Labor Economics, 15, 3, 318–37.Kletzer, L. (2000), ‘Trade and Job Loss in U.S. Manufacturing, 1979–94’, in R. Feenstra (ed.), The

Impact of International Trade on Wages (Chicago: University of Chicago Press), 349–96.Knetter, M. (1993), ‘International Comparisons of Pricing-to Market Behavior’, American Economic

Review, 83, 3, 473–86.Lemieux, T., W. B. MacLeod and D. Parent (2009) ‘Performance Pay and Wage Inequality’, Quarterly

Journal of Economics, 124, 1, 1–49.Nickell, S. and S. Wadhwani (1990), ‘Insider Forces and Wage Determination’, Economic Journal, 100,

401, 496–509.Nucci, F. and A. F. Pozzolo (2001), ‘Investment and the Exchange Rate: An Analysis with Firm-level

Panel Data’, European Economic Review, 45, 259–83.Nucci, F. and A. F. Pozzolo (2010), ‘Exchange Rate, Employment and Hours: What Firm-level Data

Say’, Journal of International Economics, 82, 1, 112–23.



Nucci, F. and A. F. Pozzolo (2012), ‘Exchange Rate, External Orientation of Firms and Wage Adjust-ment’, Sapienza University of Rome, DSS Empirical Economics and Econometrics Working PaperSeries, 3.

Organization for Economic Cooperation and Development (OECD) (2004), OECD Employment Outlook(Paris: OECD).

Revenga, A. (1992), ‘Exporting Jobs? The Impact of Import Competition on Employment and Wages inUS Manufacturing’, Quarterly Journal of Economics, 94, 1, S111–43.

Roodman, D. (2009), ‘How to Do xtabond2? An Introduction to "Difference" and "System" GMM inStata’, Stata Journal, 9, 1, 86–136.

Vandenbussche, H. and J. Konings (1998), ‘Globalisation and the Effects of National versus Interna-tional Competition on the Labour Market: Theory and Evidence from Belgian Firm Level Data’,The World Economy, 21, 8, 1151–77.

Voigtlnder, N. (2013), ‘Skill Bias Magnified: Intersectoral Linkages and White-collar Labor Demand inUS Manufacturing no access’, The Review of Economics and Statistics, doi: 10.1162/REST_a_00390.

Yang, J. (1997), ‘Exchange Rate Pass-through in US Manufacturing Industries’, Review of Economicsand Statistics, 79, 1, 95–104.



Exchange Rate, External Orientation of Firms and Wage Adjustment

Documents

Transcript of Exchange Rate, External Orientation of Firms and Wage Adjustment