Skill Upgrading and the Real Exchange Rate

The World Economy (2009) doi: 10.1111/j.1467-9701.2009.01203.x Skill Upgrading and the Real Exchange Rate SKILL World TWEC Original Oxford, XXX ROBERTO © 1467-9701 0378-5920 2009Economy UPGRADING The UK Article ÁLVAREZ Author Journal AND ANDTHE RICARDO compilation REAL EXCHANGE A.©LÓPEZ Blackwell RATE Publishers Ltd. 2009 Blackwell Publishing Ltd Roberto Álvarez1 and Ricardo A. López2 1 Central Bank of Chile and University of Chile and 2Indiana University 1. INTRODUCTION NE of the salient characteristics of many developing countries, especially those from Latin America and Africa, is the high degree of income inequality between skilled and unskilled workers. Recent changes in trade orientation among these countries have provided interesting cases for studying how inequality responds to trade liberalisation. Most of the empirical evidence in this area shows that trade liberalisation is usually accompanied by increases in wage inequality (Goldberg and Pavcnik, 2004, 2007).1 This phenomenon is puzzling because developing countries are likely to be abundant in unskilled labour. According to the Stolper–Samuelson theorem, once these countries open to trade, the returns to unskilled workers should increase in absolute terms as well as relative to the returns of skilled workers. Scholars have tried to explain this counterintuitive result from different angles. Some have argued that this pattern is consistent with the idea that developing countries are relatively abundant in natural resources rather than in unskilled labour (Leamer et al., 1999). Others have claimed that trade liberalisation has been accompanied by skill-biased technological change, O The authors would like to thank Sarah Guillou, Stefanie Haller, Richard Kneller, two anonymous referees, and participants at the Fall 2008 Midwest International Economics Meetings at Ohio State University and the International Activities and Firm Performance Workshop at the University of Nottingham, Nottingham, UK, in February 2008 for many helpful comments and suggestions. 1 One exception is the case of Brazil, where wage inequality decreased during the trade liberalisation (Gonzaga et al., 2006). For other country-specific studies on this relationship, see Beyer et al. (1999), Hanson and Harrison (1999), Galiani and Sanguinetti (2003), Attanasio et al. (2004) and Acosta and Montes-Rojas (2008). © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA 1165 1166 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ which has increased the demand for skilled workers relative to unskilled workers (Robbins, 1996; Tokarick, 2005; Gallego, 2006).2 While most of the empirical studies focus on the inequality effects of trade liberalisation, little work exists examining the effects of changes in export profitability on wage inequality.3 In particular, not much attention has been given to exchange rate fluctuations and their impact on skill upgrading and wage inequality.4 The purpose of this study is to fill this gap. This paper complements previous evidence on trade and inequality by studying how changes in the real exchange rate (RER) affect wage inequality using plant-level data. We use information on Chilean manufacturing plants for the period 1990–99 and compute three-digit-level industry-specific RERs. Thus, we extend Verhoogen’s (2008) paper that uses aggregate exchange rate fluctuations in Mexico.5 We also try to identify one possible mechanism by which the RER and wages are related.6 This question is important for several reasons. First, episodes of significant trade reforms are becoming less common as many countries have already liberalised their economies. But as countries trade more with one another they are exposed to other aspects of globalisation such as changes in exchange rates. Thus, an understanding of the impact of globalisation requires an examination of the effects of exchange rate movements on skill upgrading. Second, changes in the exchange rate may have similar effects as changes in tariffs on the price of imports, as shown by Feenstra (1989).7 Third, recent models of international trade with heterogeneous firms suggest that a real depreciation may induce skill upgrading, and contribute to increase wage inequality, by inducing firms to adopt new technologies and upgrade the quality of their products (Yeaple, 2005; Verhoogen, 2008; López, 2009). 2 The idea is that liberalising developing countries may adopt skill-biased technologies developed in industrialised nations and this technology adoption may be motivated by tariff reductions in developing countries (as shown by Attanasio et al., 2004, for the case of Colombia). Zhu (2004, 2005) provides an alternative explanation based on product cycles. She argues that the reallocation of product-cycle goods from developed countries to developing countries increases the relative demand for skilled workers in developing countries, which raises wage inequality. 3 Changes in export profitability may be due to reductions in (fixed and variable) trade costs of exporting, such as free-trade agreements or decreases in transport costs. 4 One of the few papers examining the role of changes in the real exchange rate on income inequality is Guillaumont Jeanneney and Hua (2001), which shows that a real depreciation increased income inequality between urban and rural areas in China. 5 Robertson (2003) also explores the relationship between aggregate exchange rate and relative wages in Mexico. 6 Some scholars have used industry-specific exchange rates to examine the effect of exchange rate movements on labour market outcomes (e.g. Goldberg et al., 1999), but they have not given much attention to the impact on skill upgrading. 7 This idea is called the ‘symmetry hypothesis’. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1167 This paper shows that firms facing RER depreciation experience an increase in the share of skilled labour in the total wage bill, which is consistent with models in which an increase in export profitability induces the use of more skill-intensive techniques. Our findings indicate that RER-induced skillbiased technological change within firms may explain an important part of increases in wage inequality in developing countries. The paper also shows that the effect of the RER on skill upgrading occurs through the effect on exporters. Thus, we are able to identify one possible channel by which the RER affects skill upgrading. In particular, we show that exchange rate fluctuations have significant effects on the share of foreign sales in exporting firms but do not affect the probability of exporting. Our results suggest that the intensive margin of exports is relatively more important to evaluate how real depreciations increase exports and accelerate the adoption of new technologies and product quality upgrading. The results of this paper have implications for policy. The growing participation of developing countries in world trade challenges the view that they could reduce their levels of income inequality, unless they implement specific policies to compensate those less favoured with export opportunities for the lower earnings. This paper shows that exchange rate fluctuations may also constitute a variable that policy-makers should take into account. An important issue – though not addressed directly in this paper – is where displaced unskilled labour is reallocated. In some cases the transition may not be dramatic whenever reallocation is rapid and not costly. Some countries, however, have more rigid labour markets and natural or policy-induced barriers to migration, and they may not be able to smooth this transition. In this case a rise in unemployment may occur. This paper is structured as follows. In the next section we describe the data and the main patterns. The third section presents the methodology. In the fourth section we present and discuss our results. Finally, the conclusions are presented in the fifth section. 2. DATA AND BASIC PATTERNS The plant-level data were obtained from the Annual National Industrial Survey (ENIA) carried out by the National Institute of Statistics of Chile. The ENIA contains information on sales, output, employment, wages, exports, foreign ownership, and other plant characteristics for each manufacturing plant with at least 10 employees. Each plant has a unique identification number which allows them to be followed over time. In addition, plants are classified according to the International Standard Industrial Classification (ISIC) rev. 2. This paper uses information for the years 1990–99. All monetary variables were converted to constant pesos of 1985 using four-digit ISIC-level price deflators. Data on © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1168 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ capital stock were not available so it was necessary to construct this variable using the perpetual inventory method for each plant.8 Although the ENIA distinguishes between non-production and production workers, it does not have information about the educational level or the years of experience of each worker employed in each plant. In this paper we use nonproduction workers as a proxy for skilled labour. For unskilled labour we use information on production workers. As shown by Slaughter (2000), the results of using measures of production and non-production workers are comparable to those using levels of education as measures of skill, which gives us confidence that our results can be interpreted in the context of skill upgrading and wage inequality between skilled and unskilled workers. The share of skilled labour in the total wage bill for the manufacturing sector in Chile increased from 51 per cent in 1990 to 57 per cent in 1999. The share stayed around 50 per cent between 1990 and 1995 but then it started to increase steadily from 1996. The importance of skilled labour in total wages differs across sectors. Sectors such as chemicals and machinery and equipment have average shares for 1990–99 above 50 per cent and in some cases above 60 per cent (e.g. other chemical products). Sectors with low shares include footwear, wood products, furniture and ceramics. The rate of change in the share of skilled labour also differs across sectors. Most sectors experienced an increase in the share of skilled labour between 1990 and 1999, with the highest increase occurring in professional equipment, apparel, non-ferrous metals and printing. Sectors such as electrical machinery, food miscellaneous, plastics and furniture experienced small decreases in the share of skilled labour. The increase in the share of skilled labour in total wages suggests that plants may have experienced skill upgrading during the period. It is possible that external factors may have contributed to this either directly or indirectly. Factors such as foreign technology acquisition and links with multinational corporations constitute direct sources of technology and may induce skill upgrading of local plants. In the case of Chile these sources are present in a relatively low fraction of plants. During the period 1990–99 an average of 5.6 per cent of plants purchased foreign technologies through licensing, while only 6 per cent of plants had some foreign ownership. But indirect sources of foreign technology may also be important determinants of skill upgrading in Chilean manufacturing. Contacts with foreign customers through exporting, and with suppliers through importing intermediate inputs, may allow plants to acquire information about foreign technologies, which may induce skill upgrading. A significant fraction of Chilean manufacturing plants participate in 8 For the majority of plants, an initial value of the capital stock was available. This initial value was used to construct the capital stock data by adding investment and subtracting depreciation for each type of capital (machinery and equipment, buildings and vehicles). For a small group of plants it was not possible to construct the stock of capital, so they were dropped from the dataset. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1169 international markets either through exporting or importing intermediate inputs. More than 22 per cent of the plants exported part of their output during 1990–99, while more than 25 per cent imported intermediate inputs during the same period. The fraction of plants in both categories increased during the period, suggesting that foreign influences through exporting and importing may have become increasingly important. 3. METHODOLOGY We follow the methodology used by Pavcnik (2003) where the share of skilled labour in the total wage bill is derived from a translog cost function. This approach assumes that capital is a quasi-fixed factor of production and that plants minimise the cost of skilled and unskilled labour. This cost minimisation yields the following expression: Share = β0 + β1 ln(ws/w u) + β2 ln(K/ Y ) + β3 ln(Y ) + β4Tech + ε, (1) where Share is the share of skilled labour in the total wage bill, w s is the skilled labour wage, w u is the unskilled labour wage, Y is value added, K is capital, Tech represents the observed component of technology, and ε is the unobserved technological component. We expand this equation in two main ways. First, we proxy the unobserved technological component not only by the value of foreign technology licences as in Pavcnik (2003), but also by the value of imported intermediate inputs and foreign ownership. Second, we explore how industry-specific changes in export profitability may affect skill upgrading. To do that, we compute three-digit ISIC sector-level real exchange rates. The natural log of the real exchange rate for industry j at time t is calculated as: ⎞ ⎛ C Log(RERjt ) = Log⎜ ∑α cj RERct ⎟ , ⎠ ⎝ c=1 where RERct is the bilateral real exchange rate between Chile and country c;9 C = 15 is the number of countries; and αcj is defined as: 9 The bilateral real exchange rate between Chile and country c is: RERct = NomERct * Pct /PChile,t. NomERct is the nominal exchange rate between Chile and country c (Chilean pesos/country’s c currency), while Pct and PChile,t are producer price-level indices for country c and Chile, respectively. The nominal exchange rates and producer prices were obtained from the International Financial Statistics of the International Monetary Fund. In cases in which the producer price was not available, the consumer price index was used. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1170 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ T α cj = 1 Exportscjt , ∑ T t=1 Exports jt where Exportscjt is the value of exports from industry j to country c at time t; Exportsjt is the value of exports from industry j at time t; and T is the number of periods (nine years).10 In other words, the real exchange rate is a weighted average of bilateral real exchange rate indices between Chile and the 15 main countries of destination of Chilean exports for each industry. These countries represent between 81.2 per cent and 99.5 per cent of total exports in each sector. The average coverage across all sectors is 92 per cent. An increase in this variable represents a real depreciation of the Chilean currency. Most of the manufacturing industries experienced a real depreciation between 1990 and 1999. The most important increases corresponded to printing and transport equipment. By contrast, some other industries, such as food and paper, saw reductions in their real exchange rates. Using plant-level data, we estimate the following equation: Shareijt = δi + δ1 ln(Kijt−1/ Yijt) + δ2 ln(Yijt) + δ3 Licijt + δ4 Impijt + δ5 Forijt + δ6 ln(RERjt) + γ Yeart + εijt, (2) where i denotes a plant, j a three-digit ISIC industry, Year is a vector of year-specific effects, and δi are plant-specific fixed effects. Lic, Imp and For are measures of foreign technology licensing, imports of intermediate inputs and foreign ownership, respectively (see the next section for the exact way in which we measure them). Since the current capital stock may be affected by relative wages, the regressions use the lagged value of capital stock instead of the current value. Note that equation (2) differs from equation (1) because it excludes relative wages. Pavcnik (2003) argues in favour of this practice because most of the variation in relative wages across plants is endogenous. We also try to avoid this problem by excluding this variable. As a robustness check, we estimate (2) including the relative wage between skilled and unskilled workers. The estimates for all other variables remain unchanged.11 10 We use the average over time to isolate the effect of changes in relative prices rather than changes in export shares. 11 We also estimate the equation by using the median of relative wages in the region where the plant is located instead of plant-specific relative wages. The results, which are available from the authors upon request, do not change. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1171 4. RESULTS Before presenting the estimation results of equation (2), we first investigate the relationship between the RER and exporting activity. Verhoogen (2008) and López (2009) argue that a real depreciation may potentially affect product quality and skill upgrading by increasing the probability of exporting and/or the export intensity of existing exporters. a. The RER and Exporting Activity We assume that plants make two decisions with respect to exporting. First, they decide whether or not to export. And second, they decide how much to export. Thus, plants must pass two hurdles before they are observed with a positive level of exports. We examine the role of RER on exporting by estimating the double-hurdle model developed by Cragg (1971). One advantage of this model is that it allows zero values for exports to be the result either of censoring, faulty reporting or some random event. As explanatory variables we include the RER and a set of plant characteristics such as total factor productivity (the natural log), size (the log of the number of employees), age (in log), a dummy for plants with foreign ownership, a dummy for plants that use foreign technology licences, and a dummy for plants that use imported intermediate inputs. The regressions also include sector and year fixed effects. The results are presented in Table 1. The first column shows the estimates for export participation, which is measured as a dummy variable equal to 1 if the plant exported at time t. The second column shows the estimates for export intensity (exports/total sales). We include all plant characteristics lagged one period and estimate both models (export participation and export intensity) simultaneously. Consistent with previous studies, we find that more productive and larger plants are more likely to export.12 Plants with foreign ownership and those that use foreign licences and imported intermediate inputs are also more likely to export. Age is negatively correlated with the probability of exporting, but this may be due to a non-linear relationship between age and the probability of exporting. In column (3) we add the square of age and also the squared of employment to see whether there is a non-linear relationship for size and age. The estimate for employment is still positive but for the squared term it is negative, suggesting that larger plants are more likely to export but at some level of employment the probability starts to decrease. For age we observe a similar pattern. Older plants are more likely to export but at some point the probability of exporting decreases. 12 See, for example, Bernard and Jensen (2004). © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1172 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ TABLE 1 The RER and Exporting Activity – Double Hurdle Model Log(TFP) Log(Employment) (1) (2) Export Dummy Exports/Sales (3) Export Dummy (4) Exports/Sales 0.105 (4.86)** 0.872 (36.59)** 0.000 (0.06) 0.078 (8.15)** 0.480 (4.42)** 0.255 (3.16)** 0.405 (4.32)** −0.336 (7.44)** 0.089 (8.14)** 0.017 (1.47) 0.082 (4.66)** −0.027 (4.72)** 0.326 (1.07) 0.086 (2.25)* 0.087 (4.24)** 1.924 (12.14)** −0.127 (7.30)** 0.383 (4.43)** 0.208 (3.11)** 0.392 (5.70)** 0.289 (2.45)* −0.206 (5.99)** 0.329 (0.75) 0.005 (0.92) 0.309 (8.89)** −0.024 (6.12)** 0.087 (8.85)** 0.027 (2.41)* 0.091 (7.95)** 0.008 (0.51) −0.013 (2.80)** 0.096 (2.71)** Log(Employment) Squared Foreign Ownership Foreign Technology Licences Importer Intermediate Inputs Log(Age) Log(Age) Squared Log(RER) Observations Log Likelihood 32,126 −9,305.292 32,126 −9,150.862 Notes: Robust z-statistics in parentheses. * Significant at 5%; ** significant at 1%. Standard errors clustered at the threedigit ISIC sector and year. Regressions include sector and year fixed effects. The estimate for the RER is positive but not statistically significant. This result is consistent with previous studies (e.g. Álvarez and López, 2008; López, 2009), and it is probably due to hysteresis in export markets. The hysteresis effect is based on the existence of sunk-entry cost into foreign markets.13 If these costs are significant, a small real depreciation may not induce entry into exporting. Only large and persistent fluctuations in real exchange rates should induce changes in entry and exit decisions. Columns (2) and (4) show the estimates for the export intensity. We find that larger plants, those with foreign ownership, and plants that use imported intermediate inputs export a larger fraction of their total sales. When adding squared terms for employment and age we see that the largest and oldest plants tend to have lower export intensity. The estimate for the RER is positive and 13 These sunk-entry costs of exporting have been documented by Bernard and Jensen (2004) and Das et al. (2007). © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1173 TABLE 2 Share of Skilled Wages in Total Wage Bill – Basic Regressions (1) (2) (3) (4) (5) (6) All Plants Non-exporters Exporters All Plants Non-exporters Exporters Log(Capital/Value Added) Log(Value Added) Foreign Ownership Foreign Technology Licences Importer Intermediate Inputs Log(RER) Foreign Licences/Sales Imported Inputs/ Total Inputs Constant Observations R-squared (within) −0.002 (1.41) −0.005 (2.60)** 0.007 (0.98) 0.008 (2.08)* 0.006 (2.08)* 0.025 (2.31)* −0.002 (1.25) −0.005 (2.27)* −0.027 (2.21)* 0.001 (0.17) 0.003 (0.91) 0.017 (1.27) 0.353 (6.84)** 0.310 (5.20)** 31,958 0.030 24,024 0.028 −0.001 −0.002 (0.18) (1.29) −0.006 −0.005 (1.36) (2.42)* 0.025 0.008 (2.93)** (1.07) 0.013 (1.98)* 0.003 (0.56) 0.072 0.026 (4.49)** (2.37)* 0.263 (2.09)* 0.018 (3.98)** 0.243 0.349 (2.53)* (6.82)** 7,934 0.033 31,958 0.030 −0.002 (1.25) −0.005 (2.36)* −0.027 (2.20)* −0.001 (0.20) −0.006 (1.33) 0.025 (2.94)** 0.017 (1.27) 0.106 (0.65) 0.014 (2.17)* 0.310 (5.19)** 0.071 (4.44)** 0.616 (3.17)** 0.012 (1.56) 0.247 (2.59)* 24,024 0.028 7,934 0.033 Notes: Robust t-statistics in parentheses. * Significant at 5%; ** significant at 1%. Standard errors were clustered at the three-digit ISIC sector and year. Regressions include plant and year fixed effects. statistically significant, implying that real depreciations increase export intensity among exporters.14 In summary, these results provide evidence that real depreciations increase export intensity but do not induce entry to export markets. If goods for export markets are relatively skill intensive, then a real depreciation that increases export intensity may induce firms to hire more skilled labour, so that the share of skilled wages in the total wage bill increases. b. Skill Upgrading and the RER The basic results for the effect of the RER on the share of skilled labour are presented in Table 2. We show estimates using the sample of all plants, as well 14 This result is consistent with Greenaway et al. (2007), who show that a one-percentage-point appreciation of the industry-specific RER in the United Kingdom reduces firms’ export share by 1.28 per cent. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1174 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ as for non-exporters and exporters separately. If changes in RER induce exporters to upgrade the quality of their products, then it is likely that the RER affects the share of skilled labour mostly on exporters. We use two alternative ways to measure the influence of foreign technology licences and imports of intermediate inputs. The first simply uses a dummy variable equal to 1 for plants that report the use of foreign technology licences and imported inputs. The results of the estimations using these variables are shown in columns (1) – (3). The second alternative takes into account the intensity of foreign licences and importing inputs in each plant. This is done by using the value of royalties and licence fees divided by sales as a measure of the intensity of foreign technology acquisition, and by using the ratio of imported intermediate inputs over total intermediate inputs purchased by the plant to measure the degree of reliance on foreign inputs in each plant. As seen in the table, the estimates for the ratio capital–value added are never statistically significant, suggesting that at least during this period capital is not complementary to skilled labour. In other words, within a plant, and holding everything else constant, an increase in the amount of capital a plant employs does not increase the share of skilled labour in total wages. The estimates for value added are negative but not significant for the case of exporters. In terms of the technology variables, we can see in column (1), which uses data on all plants, that the use of foreign technology licences and imported intermediate inputs increases the share of skilled labour in total wages. The dummy for foreign ownership is positive but not significant. These results suggest that foreign technology licences and importing inputs make plants more skill intensive. When distinguishing between non-exporters (column (2)) and exporters (column (3)) we see that the dummy for foreign ownership is negative in the case of non-exporters but positive for exporters, implying that links with multinational corporations increase skill intensity only if the plant participates in export markets. The estimates for the foreign technology licences dummy and the importer dummy are not significant for non-exporters, while only the dummy for foreign licences is significant in the case of exporters. When we use the intensity of licensing as well as the intensity of importing inputs we get a similar picture. Both technology variables are correlated with skill upgrading, while foreign licences appear to be more important for exporting plants. Now, however, import intensity is significant for the case of non-exporting plants. Some of these results contrast the findings of Pavcnik (2003) for the case of Chile during the period 1979–86. Using the same type of data, she finds no significant effect of several technology variables, including imported inputs, on skilled wages when using plant fixed effects. It is possible that the different results are due to the different sample periods we study. Pavcnik (2003) studies a period in which most of the trade reform was already completed and in which the economy started to specialise in natural-resource-intensive activities closely © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1175 linked to the comparative advantages of the country. Thus, although foreign technologies may be important determinants of skill upgrading, they may not have been so important during this earlier period. In order to investigate whether different time periods can explain the different results, we estimate the regressions using the same sample period used by Pavcnik (2003). Using an older version of the Annual Survey of Manufactures covering the period 1979–98,15 we find for the period 1979– 86 results that are very similar to what Pavcnik (2003) finds. But for the period 1990 –98, we find estimates that are consistent with what we find in Table 2 (see the Appendix). Our main variable of interest is the RER. As seen in columns (1) and (4), the estimate for this variable is positive and statistically significant when all plants are included in the regressions. But a closer look reveals that this significant effect is explained by the positive and significant impact of the RER on skilled wages in exporting plants. Given that increases in the RER reflect real depreciations of the Chilean currency, we can say that a real depreciation increases skill intensity, as measured by the share of skilled wages. This result is consistent with Verhoogen’s (2008) model and implies that increases in RER may induce exporting plants to upgrade product quality and become more skill intensive. Since the RER is included in logs, its estimate suggests that a 1 per cent increase in the RER increases the share of skilled workers in total wages in exporters by 0.07 percentage points. This significant increase implies that changes in the RER are likely to have important effects on skill upgrading and wage inequality. As a robustness check we also include the change in the RER as an explanatory variable. It is possible that changes in the RER are also important in explaining skill upgrading. The results are presented in Table 3. The estimates for all the technology variables do not change significantly when the change in the RER is included. The estimate for the RER remains positive and significant for the sample of exporters but not significant for non-exporters. Finally, the estimates for the change in the RER are always negative but never significant, confirming the robustness of our basic set of results. We also estimate regressions with the change in the RER but excluding the level of the RER. The results, not reported here, show that the estimate for the change in the RER is never statistically significant, while the estimates for the other control variables do not change. As a final robustness check we estimate the basic regressions excluding two sectors for which the RER increased over 100 per cent between 1990 and 1999.16 It is possible that our results are driven by the influence of these two outliers. 15 We thank Gustavo Crespi for providing us with this version of the dataset. These sectors are printing and transport equipment. The reason why these sectors experienced such a big increase in their RER is that the destination countries of their products during this period consisted mostly of Latin American countries that experienced episodes of extremely high inflation, such as Brazil, Peru and Argentina. 16 © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1176 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ TABLE 3 Share of Skilled Wages in Total Wage Bill – Regressions with Changes in RER (1) (2) (3) (4) (5) (6) All Plants Non-exporters Exporters All Plants Non-exporters Exporters − 0.002 (1.43) − 0.005 (2.61)** Foreign Ownership 0.007 (0.97) Foreign Technology 0.008 Licences (2.06)* Importer Intermediate 0.006 Inputs (2.09)* Log(RER) 0.028 (2.40)* − 0.012 Change Log(RER) (0.92) Foreign Licences/Sales Imported Inputs/ Total Inputs Constant 0.341 (6.20)** Log(Capital/Value Added) Log(Value Added) Observations R-squared (within) 31,958 0.030 − 0.002 (1.27) − 0.005 (2.29)* − 0.027 (2.20)* 0.001 (0.15) 0.003 (0.92) 0.020 (1.42) − 0.013 (0.83) 0.300 (4.69)** 24,024 0.028 − 0.001 − 0.002 (0.19) (1.31) − 0.006 − 0.005 (1.36) (2.43)* 0.025 0.008 (2.91)** (1.07) 0.013 (1.97)* 0.003 (0.56) 0.077 0.029 (4.03)** (2.47)* − 0.018 − 0.012 (0.64) (0.96) 0.261 (2.08)* 0.018 (3.99)** 0.222 0.337 (2.05)* (6.19)** 7,934 0.033 31,958 0.030 − 0.002 (1.27) − 0.005 (2.37)* − 0.027 (2.20)* − 0.001 (0.22) − 0.006 (1.33) 0.025 (2.92)** 0.020 (1.43) − 0.013 (0.84) 0.104 (0.63) 0.014 (2.17)* 0.296 (4.68)** 0.076 (3.98)** − 0.018 (0.66) 0.615 (3.16)** 0.012 (1.56) 0.226 (2.10)* 24,024 0.029 7,934 0.033 Notes: Robust t-statistics in parentheses. * Significant at 5%; ** significant at 1%. Standard errors were clustered at the three-digit ISIC sector and year. Regressions include plant and year fixed effects. The results, not presented here, show that the estimate for the RER continues to be positive and statistically significant for exporters while it is not significant for non-exporters, confirming that our basic results are robust to the exclusion of these two sectors. 5. CONCLUSIONS Recent models of international trade show that changes in export profitability affect firms’ decisions on the use of technologies and quality upgrading. This may change the employment composition of plants towards more skilled labour, thus increasing income inequality. Complementing the previous literature, this paper examines changes in industry-specific real exchange rates rather than trade liberalisation episodes. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1177 In general, the results are consistent with the idea that a real depreciation may be accompanied by an increase in the share of skilled labour in total wages. We also try to look at some of the mechanisms behind these results. We find that depreciations increase the export intensity of exporters but not the probability of exporting. Then, our results suggest that the intensive margin of exports is relatively more important to evaluate how real depreciations increase exports and accelerate the adoption of new technologies and product quality upgrading. The higher wage inequality attributable to changes in export profitability generates important questions for policy-makers in developing countries. First, it gives some evidence to evaluate the distributional effect of real exchange rate fluctuations. This is very relevant for some countries that have liberalised their trade and do not face significant trade restrictions in export markets, but they still face inequality pressures stemming from variations in the real exchange rate. Second, the growing participation of these countries in world trade challenges the view that they could reduce their levels of income inequality, unless they implement specific policies to compensate those less favoured with export opportunities for their lower earnings. A third issue – though not addressed directly in this paper – is where displaced unskilled labour is reallocated. In some cases the transition may not be dramatic whenever reallocation is rapid and not costly. Some countries, however, have more rigid labour markets and natural or policy-induced barriers to migration, and they may not be able to smooth this transition. In this case a rise in unemployment may occur. APPENDIX Pavcnik Regressions: Comparison 1979–86 and 1990–98 Log(Capital/Value Added) Log(Value Added) Foreign Technology Licences Importer Intermediate Inputs Constant Observations 1979–86 1990–98 0.007 (2.57)* 0.005 (1.62) 0.005 (1.06) −0.003 (0.92) 0.247 (8.34)** −0.008 (3.07)** −0.016 (5.03)** 0.003 (0.48) 0.005 (1.38) 0.612 (17.05)** 35,867 42,893 Notes: Robust t-statistics in parentheses. * Significant at 5%; ** significant at 1%. Regressions include plant and year fixed effects. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 1178 ROBERTO ÁLVAREZ AND RICARDO A. LÓPEZ REFERENCES Acosta, P. and G. V. Montes-Rojas (2008), ‘Trade Reform and Inequality: The Case of Mexico and Argentina in the 1990s’, The World Economy, 31, 6, 763–80. Álvarez, R. and R. A. López (2008), ‘Entry and Exit in International Markets: Evidence from Chilean Data’, Review of International Economics, 16, 4, 692–708. Attanasio, O., P. K. Goldberg and N. Pavcnik (2004), ‘Trade Reforms and Wage Inequality in Colombia’, Journal of Development Economics, 74, 2, 331–66. Bernard, A. B. and J. B. Jensen (2004), ‘Why Some Firms Export’, Review of Economics and Statistics, 86, 2, 561–69. Beyer, H., P. Rojas and R. Vergara (1999), ‘Trade Liberalization and Wage Inequality’, Journal of Development Economics, 59, 1, 103–23. Cragg, J. G. (1971), ‘Some Statistical Models for Limited Dependent Variables with Application to the Demand for Durable Goods’, Econometrica, 39, 5, 829–44. Das, S., M. J. Roberts and J. R. Tybout (2007), ‘Market Entry Costs, Producer Heterogeneity, and Export Dynamics’, Econometrica, 75, 3, 837–73. Feenstra, R. C. (1989), ‘Symmetric Pass-through of Tariffs and Exchange Rates under Imperfect Competition: An Empirical Test’, Journal of International Economics, 27, 1–2, 25–45. Galiani, S. and P. Sanguinetti (2003), ‘The Impact of Trade Liberalization on Wage Inequality: Evidence from Argentina’, Journal of Development Economics, 72, 2, 497–513. Gallego, F. (2006), ‘Skill Premium in Chile: Studying the Skill Bias Technical Change Hypothesis in the South’, Central Bank of Chile Working Paper No. 363. Goldberg, P. K. and N. Pavcnik (2004), ‘Trade, Inequality, and Poverty: What Do We Know? Evidence from Recent Trade Liberalization Episodes in Developing Countries’, Brookings Trade Forum 2004, 223 – 69. Goldberg, P. K. and N. Pavcnik (2007), ‘Distributional Effects of Globalization in Developing Countries’, Journal of Economic Literature, 45, 1, 39–82. Goldberg, L., J. Tracy and S. Aaronson (1999), ‘Exchange Rates and Employment Instability: Evidence from Matched CPS Data’, American Economic Review Papers and Proceedings, 89, 2, 204 –10. Gonzaga, G., N. Menezes-Filho and C. Terra (2006), ‘Trade Liberalization and the Evolution of Skill Earnings in Brazil’, Journal of International Economics, 68, 2, 345–67. Greenaway, D., R. Kneller and X. Zhang (2007), ‘Exchange Rates and Exports: Evidence from Manufacturing Firms in the UK’, GEP Research Paper No. 2007/13, University of Nottingham. Guillaumont Jeanneney, S. and P. Hua (2001), ‘How does the Real Exchange Rate Influence Income Inequality between Urban and Rural Areas in China?’, Journal of Development Economics, 64, 2, 529– 45. Hanson, G. and A. Harrison (1999), ‘Trade Liberalization and Wage Inequality in Mexico’, Industrial and Labor Relations Review, 52, 2, 271–88. Leamer, E. E., H. Maul, S. Rodriguez and P. Schott (1999), ‘Does Natural Resource Abundance Increase Latin American Income Inequality?’, Journal of Development Economics, 59, 1, 3 –42. López, R. A. (2009), ‘Do Firms Increase Productivity in Order to Become Exporters?’, Oxford Bulletin of Economics and Statistics (forthcoming). Pavcnik, N. (2003), ‘What Explains Skill Upgrading in Less Developed Countries?’, Journal of Development Economics, 71, 2, 311–28. Robbins, D. (1996), ‘HOS Hits Facts: Facts Win; Evidence on Trade and Wages in the Developing World’, Development Discussion Paper No. 484 (Cambridge, MA: Harvard Institute for International Development). Robertson, R. (2003), ‘Exchange Rates and Relative Wages: Evidence from Mexico’, North American Journal of Economics and Finance, 14, 1, 25–48. Slaughter, M. J. (2000), ‘What are the Results of Product Price Studies and what can we Learn from their Differences?’ in R. C. Feenstra (ed.), The Impact of International Trade on Wages, NBER Conference Volume (Chicago: University of Chicago Press), pp. 129–70. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 SKILL UPGRADING AND THE REAL EXCHANGE RATE 1179 Tokarick, S. (2005), ‘Quantifying the Impact of Trade on Wages: The Role of Nontraded Goods’, Review of International Economics, 13, 5, 841–60. Verhoogen, E. (2008), ‘Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing Sector’, Quarterly Journal of Economics, 123, 2, 489–530. Yeaple, S. (2005), ‘A Simple Model of Firm Heterogeneity, International Trade, and Wages’, Journal of International Economics, 65, 1, 1–20. Zhu, S. C. (2004), ‘Trade, Product Cycles and Inequality Within and Between Countries’, Canadian Journal of Economics, 37, 4, 1042–60. Zhu, S. C. (2005), ‘Can Product Cycles Explain Skill Upgrading?’, Journal of International Economics, 66, 1, 131–55. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009