WPS7676 Policy Research Working Paper 7676 Exports and Job Training Paulo Bastos Joana Silva Rafael Proença Development Research Group Trade and International Integration Team & Latin America and the Caribbean Region Office of the Chief Economist May 2016 Policy Research Working Paper 7676 Abstract This paper examines whether export participation mat- participation tends to increase the share of workers who ters for job training. The paper draws on longitudinal receive technical upgrading. The results also reveal that worker-firm data for Brazilian manufacturing, linked with technical upgrading has positive returns to trainees within detailed records on training activity from the main pro- exporting firms. These findings support the hypothesis that vider. The analysis uses industry-specific exchange rate exporting requires skill upgrading, and suggest that this movements to generate exogenous variation in export is partially achieved by training firms’ existing workforce. status at the firm-level. The findings indicate that export This paper is a product of the Trade and International Integration Team, Development Research Group, and the Office of the Chief Economist, Latin America and the Caribbean Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at pbastos@ worldbank.org, jsilva@worldbank.org, or rproenca@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Exports and Job Training∗ Paulo Bastos† Joana Silva‡ ca§ Rafael Proen¸ Keywords : Export participation, skill upgrading, job training, wages, linked employer-employee data. JEL Codes : F12, F14, F16, F66, J24, J31, J38, O14. ∗ We are grateful to Aguinaldo Maciente, Alejandra Martinez, Carlos Silva dos Santos, Daniel Lederman, ao, Laura Chioda, Renata Gukovas, Renato da Fonseca, Rita Almeida and participants at Herlon Brand˜ the Festchrift Conference in Honor of Professor Sir David Greenaway at the University of Nottingham for helpful comments and discussions. We are also grateful to the Ministry of Labor, the Ministry of Development, Industry and Foreign Trade and the National Confederation of Industry (CNI) for providing the administrative data used in the paper. Research for this paper has been supported in part by the World Bank’s Multidonor Trust Fund for Trade and Development and through the Strategic Research Partnership on Economic Development. We remain responsible for any errors. † The World Bank. E-mail: pbastos@worldbank.org. ‡ Corresponding author: The World Bank, 1818 H Street NW, Mail Stop I9-900, Washington DC, United States. E-mail: jsilva@worldbank.org. § The World Bank. E-mail: rproenca@worldbank.org. 1 Introduction A growing body of literature suggests that export participation leads to an increase in skill utilization within firms (Verhoogen 2008; Brambilla, Lederman and Porto, 2012). This relationship is consistent with several intuitive theoretical mechanisms. First, exporting may require expertise in international business and foreign languages (Matsuyama, 2007; Brambilla, Lederman and Porto, 2012). Second, it may induce firms to upgrade product quality, which is a skill intensive activity (Verhoogen, 2008; Bastos and Silva, 2010; Kugler and Verhoogen, 2012; Brambilla, Lederman and Porto, 2012; Bastos, Silva and Verhoogen, 2014). Third, by leading to an expansion in the scale of operations, entry in export markets may make it profitable for firms to pay the fixed costs associated with the adoption of more skill-intensive technologies (Yeaple, 2005; Bustos 2011). While previous empirical research has documented a robust causal effect of export participation on skill intensity—as measured by workers’ schooling or occupation—, little is known about whether it leads to skill upgrading of the existing workforce through job training. Since skilled labor is a scarce factor and skill-acquisition is costly, training may be a relevant channel of adjustment to globalization, with potentially important implications for productivity and labor earnings. In this paper, we examine these hypotheses in a rich combination of administrative data for the Brazilian manufacturing sector, linking a worker-firm labor census with customs data on export status and detailed training records from the main provider. The worker-firm data we use cover the universe of manufacturing firms formally reg- istered in Brazil, and include longitudinal records on earnings and worker attributes, as well as firm characteristics. Using unique firm identifiers, we have linked these data with information from customs declarations indicating whether the firm had positive export revenues in each year. Using unique worker identifiers, we have matched these data with detailed administrative records on training activity at the trainee-level from the main provider in the manufacturing sector. The individual training records contain detailed information on training category and course duration, enrollment and completions dates. The training courses provided are grouped in five categories, which serve different objec- 2 tives. Technical upgrading courses are aimed at workers seeking to update, deepen or complement the professional skills they have acquired in a specific occupation or field or work. In contrast, the other training categories are aimed at training a new generation of practitioners of a given trade or profession. The resulting data set gathers this information for the years 2009 to 2012, a period characterized by high volatility of the real exchange rate relative to several important export destinations. To identify the effects of export participation on the share of workers receiving different types of training within firms, we first adopt a difference-in-differences approach using the full sample of manufacturing firms. We then restrict the attention to switchers—that is, firms that recorded export activity in a subset of the sample period— thereby exploiting solely variation in the timing of exporting across firms. Controlling for observed firm attributes, results from both these approaches suggest that export partici- pation is positively associated with the share of workers who received technical upgrading, but show no effects on the share of workers who completed training in other training categories. While the focus on switchers contributes to reduce biases associated with firm hetero- geneity, one may still be concerned that unobserved firm-specific shocks (e.g. in manage- ment practices) might lead some firms to both enter export markets and seek to upgrade the technical skills of the workforce. To account for this concern, we take advantage of the fact that firms operating in different industries were exposed to movements in differ- ent currencies to generate plausibly exogenous variation in the timing of exporting at the firm-level. The IV results point once more to a positive effect of export participation on the share of workers receiving technical upgrading within firms, but again show no effects for the other training categories. For technical upgrading, the point estimates are positive across all occupational groups, but more precisely estimated among workers performing occupations that typically do not require post-compulsory formal education. While the data do not allow us to fully discriminate between the various theoretical channels dis- cussed above, these results appear to suggest that one or more of these mechanisms played an important role in driving these effects. 3 Given that export participation has a positive effect on the provision of technical upgrading within firms, an important complementary question is the extent to which the corresponding trainees benefit from these additional skills in the form of higher labor earnings. To examine this question, we exploit detailed information on training records and earnings trajectories at the worker-firm level. Because entry in export markets may affect wages via several forms of rent sharing (Egger and Kreickemeier, 2009; Egger, Egger and Kreickemeier, 2012; Davis and Harrigan, 2009; Helpman, Itskhoki and Redding, 2010), we restrict our attention to firms that were always exporters over the period of analysis. Using models with worker-firm fixed effects to account for unobserved heterogeneity of workers and firms, we document a robust positive relationship between completion of technical upgrading courses and hourly wages within exporting firms. This relationship holds within small, medium and large exporting firms. In addition to the work cited above, this paper complements and extends other strands of existing research. In particular, it adds to a growing empirical literature using worker- ıas, Kaplan and firm data to examine the link between exporting, skills and wages (Fr´ Verhoogen, 2012; Helpman, Itskhoki, Muendler and Redding, 2012). An important ques- tion in this literature has been the extent to which higher wages among exporting firms reflect sorting on individual ability or wage premia due to rent sharing. By suggesting that export participation leads to skill upgrading of firms’ existing workforce via job training, and that this training leads in turn to higher wages, we uncover a related but still dis- tinctly different channel by which exporting influences skill utilization and labor earnings. In doing so, our paper adds to a small literature on how other aspects of globalization in- fluence job training. In particular, Hogrefe and Wrona (2015) develop a theoretical model examining the impact of offshoring on individual skill upgrading through on-the-job train- ing. The model predicts that offshoring, by scaling up workers’ wages, creates previously unexploited skill upgrading possibilities and, thus, leads to more on-the-job training. Us- ing individual-level survey data from Germany, they document a positive relation between an industry-level measure of offshoring and job training.1 1 More broadly, our paper is also related to the literature on labor market adjustment to globalization; 4 The remainder of the paper is organized as follows. In section 2, we provide a detailed description of the data employed and provide summary statistics. In section 3, we present our identification strategy for examining the impacts of exporting on the share of workers trained, and present the corresponding results. In section 4, we use the worker-firm infor- mation to examine whether technical upgrading is associated with higher labor earnings for the corresponding workers within exporting firms. Section 5 concludes. 2 Data 2.1 Data description The analysis in this paper links and examines three main sets of administrative data for the c˜ Brazilian manufacturing sector in the period 2009-2012. Rela¸ c˜ ao Anual de Informa¸ oes Sociais (RAIS) is a longitudinal census gathering administrative social security records for employees and employers. It is collected by the Ministry of Labor in a compulsory survey of all firms and their registered workers, and covers a total of around 230 thousand for- mally registered firms and over 3.5 million workers each year in the manufacturing sector. RAIS provides information on workers’ demographics (age, gender, schooling, race), job characteristics (occupation, wage, hours worked), as well as hiring and termination dates. It also includes information on a number of firm-level characteristics, notably number of employees, geographical location and industry code. Unique identifiers for workers and firms make it possible to follow them over time. We have performed extensive checks to guarantee the accuracy of the worker data using information on gender, date of birth, and maximum schooling level achieved. Only full-time wage earners with complete information on demographics and job attributes were kept for the analysis.2 Using the unique firm identifiers, we have merged these data with customs records on ercio Externo (SECEX) of the Ministry export transactions collected by Secretaria do Com´ see Greenaway and Nelson (2000), Bastos and Silva (2009), Falvey, Greenaway and Silva (2008, 2010) and the references therein. 2 In particular, we kept for analysis employees working more than 20 hours per week, aged between 16 and 65, and earning at least the national minimum wage. 5 of Development, Industry and Foreign Trade. These data include information on firm-level export status in each year over 2009-2012, as well as the share of industry-level exports to each destination in 2008. To construct industry-specific real exchange rate instruments, we have supplemented these data with information on the consumer price index from the World Development Indicators of the World Bank and bilateral nominal exchange rates from the International Financial Statistics of the International Monetary Fund. We further use detailed administrative records on training provision collected by the co Nacional de Aprendizagem largest training provider in the manufacturing sector–Servi¸ Industrial (SENAI), the training arm of the national confederation of industry.3 From these records, we extracted the set of workers trained in each year, worker-level information on demographics, occupation before starting the course, training category and course duration, as well as enrollment and completion dates. The data cover around 270,000 trainees per year. SENAI provides five categories of job training: (1) technical upgrading; (2) apprentice- ship; (3) initiation; (4) habilitation; and (5) qualification courses. The various course types have different objectives. In particular, technical upgrading courses differs considerably from the other training categories. Technical upgrading courses are aimed at workers seeking to update, expand or com- plement the professional skills they have acquired in a specific occupation or field of work. Their duration varies between 40 and 120 hours. The curriculum is advanced and occupation-specific. They focus mainly on the skill needs resulting from technological innovations and new production and management processes. For instance, for industrial electricians, SENAI offers a course on electrical controls which lasts 80 hours. 3 SENAI is the national service for training in the Brazilian industrial sector. It consists of a network of ca professional schools established and maintained by Confedera¸ ustria (CNI), the national ˜o Nacional da Ind´ confederation of industry. SENAI is part of Sistema S which is a set of vocational training institutions managed by CNI. While SENAI is the manufacturing arm, CNI also manages similar institutions for other sectors, including wholesale trade, agriculture and transport. SENAI is one of the 5th largest training providers in the world and the largest in Latin America. It is financed by a tax charged to firms equal to 1% of their wage bill and is the main provider of job training in manufacturing (Silva, Almeida and Strokova, 2015). 6 In contrast, the other categories provided by SENAI are aimed at training a new generation of practitioners of a trade or profession. Apprenticeship consists of on-the-job training and often some accompanying study (classroom work and reading) to learn a trade or profession. They typically last 2 years and target the youth. Habilitation courses are also long duration courses (1,100 to 1,700 hours) targeted at high school students or graduates. They provide minimum qualifications for certain mid-level technical jobs (e.g. industrial automation), and seek to complement traditional education while the student is at school or unemployed. Qualification courses target older, low-skill workers and aim at preparing them for a new trade or profession. This category includes job retraining programs for workers who need new qualifications due to technological or organizational change. Generally it reflects changes in occupation rather than an upward movement in the same field. Courses are basic-level and duration varies between 160 and 500 hours. Initiation courses have similar objectives as qualification courses but are shorter duration (40 to 80 hours) and specialized on a certain function. For example, there is a qualification course for automotive painter job (180 hours) which includes three modules: automotive refinish (120 hours), colorimetry (40 hours) and polishing and crystallization (20 hours); and there are initiation courses for a car polisher job (40 hours) and automotive paint preparation job (80 hours). 2.2 Summary statistics Our empirical analysis uses data for the years 2009 to 2012, a period characterized by high real exchange rate volatility. Panel A of Figure 1 depicts the evolution of the real effective exchange rate. We observe that the Brazilian currency observed a marked effective real appreciation until July 2011, when it started to move in the opposite direction. Panel B displays the evolution of the Brazilian currency (the real ) relative to the the euro, the US dollar and the renminbi. We see that while the overall trend was broadly similar across these three currencies, the magnitude of relative changes varied considerably. Table 1 reports summary statistics on the attributes of manufacturing producers over the sample period, discriminating firms according to export status. We observe that the 7 main stylized facts about exporting firms reported in previous literature (Greenaway and Kneller, 2007) are also present in Brazil: exporters firms are clearly a minority, and tend to be larger, pay higher average wages and employ a more skilled workforce than the typical manufacturing firm. We also see that the number of exporters remained fairly stable over time. Table 2 presents descriptive statistics on the attributes of trainees in manufacturing firms, across the various training categories. We observe that trainees receiving technical upgrading courses are more likely to be employed before starting the course than those who have received other training categories. We also observe that they are more likely to be employed in the same firm and occupation in the year before training. The table also provides evidence on the distribution of trainees across occupational groups. We consider four occupational groups based on the 1988 International Standard Classification of Occupations (ISCO-88). This classification provides four skill-levels which are based on (i) the level of general education required to perform a job; and (ii) the job-related formal training required to perform a job (ILO, 1990).4 We observe that SENAI courses are mainly directed to workers in skill groups 2 and 1, who typically do not have post- compulsory education. Table 3 presents information on the share of workers trained in exporting and non- exporting firms by training category and occupational group. We observe that exporting firms tend to have a higher share of workers who received training. On average, the proportion of workers who completed training is 4.3% in exporting firms versus 1.5% in non-exporting firms. The share of workers who completed technical upgrading courses in exporting firms is 3.6 times higher in exporters than in non-exporters. We also observe that not all occupational groups are equally likely to have received training. In exporting firms, the share of workers trained is larger among skill groups 1 to 3. 4 See Appendix Table A1 for a detailed description. See also Maciente (2013) for a related classification. 8 3 The effect of export participation on job training 3.1 Empirical strategy We are interested in examining the causal effect of exporting on the share of workers who received job training within firms. We estimate a model of the following form: sharejt = βexportjt + γXjt + ηj + δs + θr + φt + µjt (1) where j and t index firms and year, respectively; sharejt is the share of trained workers in total employment of the firm j in year t; exportjt is the export status of the firm; Xjt are other time varying firm characteristics; ηj is a firm fixed effect; δs is a sector fixed effect, defined at the 2-digit level of the national classification of economic activities (CNAE); θr is a federal state fixed effect; φt is a year effect; and µjt is the error term. We cluster the standard errors at the firm-level. The set of firm controls include log employment, average age of workers, average tenure, male to female workers ratio, share of workers by education (less than high school, high school completed and more than high school), average log wage, share of white work- ers, share of workers by occupational groups. We compute the dependent variable using individual level information on the set of workers who completed training in each year. The inclusion of firm fixed-effects accounts for the role of unobserved firm character- istics that are constant over time. While this difference-in-differences design controls for selection into exporting based on time-invariant firm characteristics (e.g., initial productiv- ity), it does not account for the fact that exporting firms might be expected to experience different trends in the outcomes of interest relative to non-exporters. To improve compara- bility across firms, we estimate a second set of regressions where we restrict the estimation sample to include only switchers, defined as firms that recorded export activity only in a subset of years of the sample period. In this case, identification of the parameters of interest exploits solely variation in the timing of exporting across firms. A concern that remains after restricting the estimation sample to switchers is that there may be unobserved firm-specific shocks (e.g. changes in management practices) that induce some firms to both enter export markets and improve the skills of the workforce 9 through worker training. To address this concern, we use industry-specific destination- weighted real exchange-rate movements to construct an instrument for export status. Our strategy builds on the plausible assumption that an expansion of external demand in a given export destination (driven by favorable relative price movements) induces more firms to begin exporting, but especially so in industries for which that destination represents already a sizable share of total exports.5 For a given sector s and year t, the destination-weighted real exchange rate, zst , is defined as the weighted average of the (reciprocal of) bilateral real exchange rates. The weights are the share of the industry’s exports to a destination in total industry exports in 2008. Formally, we compute zst = wds2008 rpldt (2) d where wds2008 is the ratio of industry s exports to destination d over total industry exports in 2008; and rpldt is the log of the reciprocal of the corresponding bilateral real exchange rate in year t. As usually, the latter is defined as: CP Idt rpldt = log /edt (3) CP IBRt where CP Idt denotes the consumer price index of export destination d in year t, CP IBRt denotes Brazil’s consumer price index in year t, and edt the nominal exchange rate. Using the reciprocal of the real exchange rate rather than the real exchange rate itself is purely a matter of convenience: an increase in zst leads to an increase in external demand for the industry’s exports, thereby increasing the likelihood of export participation in that industry. 3.2 Results Table 4 reports the difference-in-difference estimates of the effect of export participation on the share of workers who complete training at the firm level, using the full sample 5 Similar instruments based on real exchange rates have been used at the sector-level by Revenga (1992), Bertrand (2004), Falvey, Greenaway and Silva (2008) and Bastos and Wright (2012), among others. 10 of manufacturing firms. As described in section 2, the training data include detailed information on several training categories provided by SENAI. Based on the theoretical mechanisms discussed in the introduction, technical upgrading is the training category that is more likely to be related to exporting. These are courses aimed at workers seeking to update, expand or complement the professional skills they have acquired in a specific occupation or field or work. The other categories consist of longer term investments seeking to reconvert unqualified workers or retooling for new occupations workers whose skills became obsolete. Hence, we estimate (1) separately for the share of workers who completed technical upgrading and other courses. Column 1 uses the pooled data controlling only for year effects. Column 2 adds log firm size and other firm controls, as well as state and sector dummies. Column 3 adds firm fixed effects, but excludes firm controls and state dummies. Column 4 reports results from our preferred difference-in-differences specification, which includes firm, sector and state effects, as well as firm controls. For all these specifications, the results indicate that export participation is positively associated with the share of workers receiving technical upgrading within firms. By contrast, the effects of export participation on the share of workers trained in categories other than technical upgrading are insignificant when including firm effects. Table 5 presents estimates from specifications similar to those in Table 4, but consid- ering a sample including only switchers. Once again, we find that export participation is positively associated with the share of workers who received technical upgrading. In all specifications, the coefficient of the export status is positive and statistically significant. The effects of exporting on the share of workers trained in other categories are once again insignificant. As discussed above, a concern that remains after restricting the estimation sample to switchers is that there may be unobserved firm-specific shocks (e.g. changes in manage- ment practices) that induce some firms to both enter export markets and improve the skills of the workforce through worker training. This would lead to a positive bias in OLS estimates, as we would be attributing the observed increase in technical upgrading to the effects of export participation when in reality it would be reflecting other underlying 11 forces. On the other hand, if firms anticipate that they will enter export markets in the near future, they might have an incentive to begin upgrading the skills of the workforce before actually starting to export. In this case, OLS estimates on the contemporaneous effect of export participation on training would underestimate the extent of skill upgrad- ing caused by exporting. To address these concerns, we now turn to the IV estimates on the causal effect of export participation on training. Real exchange rate movements change directly the incentives to export and are largely unpredictable. Hence they make it possible to circumvent (or at least mitigate) these challenges, and estimate the causal effect of exporting on training. Table 6 presents IV estimates of (1) based on the sample of switchers, and using the industry-level real exchange rate instruments to generate exogenous variation in the incentives to export. The upper part of the table presents the results of the first stage, while the lower panel presents the results of the second stage. Column 1 includes only firm fixed effects and column 2 adds firm controls. As expected, the first-stage estimates indicate that a real depreciation (i.e. a rise in the industry-specific exchange rate variable) leads indeed to an increase in the probability of exporting. The instrument appears to be sufficiently correlated with the exporting variable for us to be able to identify the effect of that variable on the outcomes of interest: the individual coefficient is significant at the 5% level and the F-statistic is 10.7. Estimates from the second stage point once more to a positive causal effect of export participation on the share of workers receiving technical upgrading, but not for the other training categories. Results of the instrumented variable indicate a stronger effect in this case, suggesting that the OLS estimates are biased downwards. In the analysis so far, we have examined the effects of exporting on the share of workers who completed training in that same year. In light of the theoretical mechanisms discussed in the introduction, firms might be also expected to increase the share of workers who received technical training in prior years (and are therefore already equipped with the skills required by exporting firms). While our training data do not contain information on training activity before 2009, we can use the 2009-2012 data to redefine the dependent 12 variable, so that it accounts for training activity observed in that year or in any prior year during the sample period. Table 7 reports the corresponding IV results. We observe that the results are qualitatively similar, but the coefficient on technical upgrading is larger, as would be expected. We proceed by examining whether the effect of export participation on technical up- grading is heterogeneous across occupational groups. In line with Guadalupe (2007) and Bastos, Straume and Monteiro (2009), we consider four occupational groups which require different skills (see Appendix Table A1). We then compute the share of workers who completed technical upgrading within each of these occupational groups. We adopt an IV strategy similar to that presented above, focusing on a sample of switchers that have work- ers in all four occupational groups considered. The corresponding IV results are presented in Table 8. Panels A to D report second-stage results for separate regressions each with a different dependent variable. We observe that the point estimate on the effect of export participation on technical upgrading is positive for all occupational groups, though only statistically significant for skill levels 1 and 2. It is important to note that these skill cat- egories are fairly broadly defined: they include a wide diversity of occupations performed by workers that have typically completed compulsory education, but not further education (see Table A.1 in the appendix); and account for over 80% of jobs within exporting firms (see Table 1). They also correspond to occupational groups that SENAI courses typically target. 4 Returns to training among exporters 4.1 Empirical strategy The results above indicate that export participation has a positive effect on the provision of technical upgrading within firms. A related question is the extent to which technical upgrading leads to higher labor earnings among exporting firms. To examine this hypoth- esis, we exploit detailed information on training records and earnings trajectories at the worker-firm level. Because entry in export markets may affect wages via several forms of 13 rent sharing (Egger and Kreickemeier, 2009; Egger, Egger and Kreckemeier, 2012; David and Harrigan, 2009; Helpman, Itskhoki and Redding, 2008), we restrict the attention to firms that were always exporters over the period of analysis. We estimate a model of the following form: wageijt = βtrainingit + γXit + δZjt + ηi + κj + θr + φt + µijt (4) where wageijt denotes the log of the hourly real wage of worker i employed by firm j in year t; trainingit is the main regressor of interest, a dummy variable that equals 1 if the worker received the technical upgrading (either in year t or in previous years) and 0 otherwise; Xit is a vector of worker characteristics; Zjt is a vector of firm characteristics; ηi is a worker fixed effect; κj is a firm fixed effect; θr is a state fixed effect; φt is a year effect; and µijt is an error term. Standard errors are clustered at the firm-level. The set of worker attributes includes age, gender, race, schooling and skill level based on the worker’s occupation. When including worker fixed effects, we are comparing the change in log wages of those workers who received technical upgrading (before and after training) with the change in log wages of the workers who did not receive training during the same period, thereby sweeping away unobserved worker heterogeneity that is permanent over time. To account for unobserved firm heterogeneity, we also consider specifications that include worker-firm (or spell) fixed effects.6 In this case, identification of the parameter of interest exploits variation over time in log wages and training status for workers that remain employed in the same firm, thereby sweeping away fixed unobserved heterogeneity of both workers and firms. Appendix Table A2 provides summary statistics on these data. 4.2 Results Table 9 presents the results, including a progressively larger set of controls and fixed effects. Columns 4 to 6 report the results from our preferred specifications. Column 4 includes worker controls, worker fixed effects and firm controls, but not firm fixed effects. Column 5 includes worker-firm fixed effects but excludes worker and firm controls. Finally, column 6 6 See Andrews, Schank and Upward (2006) for a detailed description of these methods. 14 includes worker-firm fixed effects, as well as worker and firm controls. The results indicate that returns to technical upgrading are positive and statistically significant in all these specifications. We also observe that the magnitude of the point estimate of interest is quite stable across these specifications, and points to positive wage effect of about 3%. While the estimation sample includes only exporting firms, one might worry that the results are driven by only a subset of those firms. We therefore proceed by verifying the extent to which these results are specific to small, medium and large exporting firms. In Table 10 we split the sample according to a firm size classification of Instituto Brasileiro ıstica (IBGE), the national statistical institute. We observe that de Geografia e Estat´ technical upgrading is associated with significantly higher labor earnings among firms in each of these categories. The sample used in Tables 9 and 10 includes among the control group workers that received other types of training. Although the analysis in the previous section suggests that export participation does not impact the share of workers trained in those other categories, we cannot exclude the possibility that such training also affects labor earnings. We therefore verify if our estimates are sensitive to the exclusion of workers that received other types of training. In particular, we re-estimate (4) on a sample that excludes all observations that received any SENAI training course other than technical upgrading during 2009-2012. Appendix Tables A3 and A4 report the corresponding results. We observe that our findings remain qualitatively and quantitatively very similar when using this alternative sample. 5 Conclusion This paper has examined the effects of export participation on job training in a rich com- bination of worker-firm panel data for the Brazilian manufacturing sector, linked with detailed records on training activity from the main provider. Using industry-specific ex- change rate movements to generate exogenous variation in export status at the firm-level, we have provided evidence that export participation tends to increase the share of workers who received training. In line with several intuitive theoretical mechanisms proposed in 15 the literature (Yeaple, 2005; Matsuyama, 2007; Verhoogen, 2008) the training category linked with exporting is technical upgrading, which is aimed at improving the skills of experienced workers for enhanced performance, rather than teaching the skills needed to enter a specific occupation or field of work (as in the other training categories). Drawing on detailed worker-firm records for exporting firms, we have also provided evidence that technical upgrading has positive returns to trainees within exporting firms. Our results therefore support the hypothesis that export participation requires skill upgrading, and suggest that this is partially achieved by training firms’ existing workforce. References Andrews, Martyn, Thorsten Schank and Richard Upward (2006): ”Practical fixed-effects estimation methods for the three-way error components model,” Stata Journal, 6: 461481. Bastos, Paulo and Joana Silva (2010): ”The quality of a firm’s exports: Where you export to matters,” Journal of International Economics, 82(2): 99-111. Bastos, Paulo and Joana Silva (2008): ”The wage and unemployment impacts of trade adjustment,”, in David Greenaway, Richard Upward and Peter W. Wright (eds.), Glob- alisation and Labour Market Adjustment, 31-53, Palgrave Macmillan. Bastos, Paulo, Joana Silva and Eric Verhoogen (2014): ”Export destinations and input prices,” NBER Working Papers 20143, National Bureau of Economic Research. alia P. Monteiro and Odd Rune Straume (2009): ”Firm heterogeneity Bastos, Paulo, Nat´ and wages in unionised labour markets: Theory and evidence,” Labour Economics, 16(4): 440-450. Bastos, Paulo and Peter Wright (2012): ”Exchange rates and wages in unionized labor markets,” Industrial and Labor Relations Review, 65(4), 975-999. 16 Bertrand, Marianne (2004): ”From the invisible handshake to the invisible hand? How import competition changes the employment relationship,” Journal of Labor Economics, 22(4): 723-765. Brambilla, Irene, Daniel Lederman and Guido Porto (2012): ”Exports, export destina- tions, and skills,” American Economic Review, 102(7): 3406-3438. Bustos, Paula (2011): ”Trade liberalization, exports, and technology upgrading: Evi- dence on the impact of MERCOSUR on Argentinian firms,” American Economic Review, 101(1): 304-340. Davis, Donald R. and James Harrigan (2011): ”Good jobs, bad jobs, and trade liberal- ization,” Journal of International Economics, 84(1): 26-36. Egger, Hartmut, Peter Egger and Udo Kreickemeier (2013): ”Trade, wages, and profits,” European Economic Review, 64(3): 332-350. Egger, Hartmut and Udo Kreickemeier (2009): ”Firm heterogeneity and the labor market effects of trade liberalization,” International Economic Review, 50(1): 187-216. Greenaway, David and Richard Kneller (2007): ”Firm heterogeneity, exporting and for- eign direct investment,” Economic Journal, 117(517): F134-F161. Greenaway, David and Douglas Nelson (2000): ”The assessment: Globalization and labour-market adjustment,” Oxford Review of Economic Policy, 16(3): 1-11. Guadalupe, Maria (2007): ”Product market competition, returns to skill, and wage in- equality,” Journal of Labor Economics, 25: 439-474. ıas, Judith, David Kaplan and Eric Verhoogen (2012): ”Exports and within-plant wage Fr´ distributions: Evidence from Mexico,” CEPR Discussion Papers 8835. Falvey, Rod, David Greenaway and Joana Silva (2010): ”Trade liberalization and human capital adjustment,” Journal of International Economics, 81(2): 230-239. 17 Falvey, Rod, David Greenaway and Joana Silva (2008): ”International competition, re- turns to skill and labour market adjustment”, GEP Research Paper 2008/10, University of Nottingham. Falvey, Rod, David Greenaway and Joana Silva (2010): ”Human capital and adjustment to trade,” in David Greenaway, Richard Upward and Peter W. Wright (eds.), Globalisa- tion and Labour Market Adjustment, 97-114, Palgrave Macmillan. Helpman, Elhanan, Oleg Itskhoki, Marc-Andreas Muendler and Stephen J. Redding (2012): ”Trade and inequality: From theory to estimation,” NBER Working Papers 17991, National Bureau of Economic Research. Helpman, Elhanan, Oleg Itskhoki and Stephen Redding (2010): ”Inequality and unem- ployment in a global economy,” Econometrica, 78(4): 1239-1283. Maciente, Aguinaldo (2013): ”The determinants of agglomeration in Brazil: input- output, labor and knowledge externalities,” PhD dissertation, University of Illinois at Urbana Champaign. Hogrefe, J. and J. Wrona (2015): ”Trade, tasks, and training: The effect of offshoring on individual skill upgrading,” Canadian Journal of Economics, forthcoming. Kugler, Maurice and Eric Verhoogen (2012): ”Prices, plant size, and product quality.” Review of Economic Studies, 79(1): 307-339. Lileeva, Alla and Daniel Trefler (2010): ”Improved access to foreign markets raises plant- level productivity... for some plants,” Quarterly Journal of Economics, 125(3): 1051-1099. ILO (1990): ISCO-88: International Standard Classification of Occupations. Interna- tional Labour Office, Geneva. Matsuyama, Kiminori (2007): ”Beyond icebergs: Toward a theory of biased globaliza- tion,” Review of Economic Studies, 74(1):237-253. 18 Revenga, Ana (1992): ”Exporting jobs? The impact of import competition on em- ployment and wages in U.S. manufacturing,” Quarterly Journal of Economics, 107(1): 255-284. Silva, Joana, Rita Almeida and Victoria Strokova (2015): ”Sustaining employment and wage gains in Brazil: A skills and jobs agenda.” Directions in Development, World Bank: Washington DC. Verhoogen, Eric (2008): ”Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector,” Quarterly Journal of Economics, 123(2): 489-530. Yeaple, S.R. (2005): ”A Simple model of firm heterogeneity, international trade, and wages,” Journal of International Economics, 65(1): 1-20. A. Appendix A.1. Definition of skill groups We consider four occupational groups to proxy the skill level associated with the worker’s occupation, as defined in the ISCO-88 classification. Table A.1 presents the definition of skill groups. A.2. Summary statistics, worker-firm data A.3. Further results 19 Table 1: Attributes of manufacturing firms by export status Non-exporters Exporters All firms Log employment 1.888 4.160 1.997 (1.231) (1.711) (1.349) Average log hourly wage 1.724 2.253 1.749 (0.371) (0.493) (0.394) Share with schooling less than high school 0.515 0.389 0.509 (0.360) (0.271) (0.357) Share with schooling equal to high school 0.437 0.451 0.438 (0.352) (0.236) (0.348) Share with schooling more than high school 0.046 0.158 0.052 (0.129) (0.178) (0.134) Share of production workers 0.352 0.366 0.353 (0.348) (0.254) (0.344) Share of workers in skill level 4 0.040 0.069 0.041 (0.135) (0.116) (0.134) Share of workers in skill level 3 0.034 0.101 0.038 (0.122) (0.126) (0.123) Share of workers in skill level 2 0.813 0.667 0.806 (0.279) (0.249) (0.280) Share of workers in skill level 1 0.111 0.161 0.113 (0.229) (0.213) (0.228) Observations 2009 224,546 11,761 236,307 2010 242,206 11,545 253,751 2011 221,807 11,331 233,138 2012 219,135 11,114 230,249 Total 907,694 45,751 953,445 Notes: Table reports means and standard deviations (in parentheses) for firms operating in the manufacturing sector over the period 2009-2012. A firm-year observation is defined as exporter if the firm had export revenue in that year. Wages are in 2010 Brazilian Reais, employment refers to the number of workers employed by the firm, and employment shares are relative to total employment. Skill levels 1 to 4 are based on the 1988 International Standard Classification of Occupations (see Appendix Table A1 for a detailed description). Table 2: Training categories and trainee characteristics in the manufacturing sector Technical Apprenticeship Habilitation Initiation Qualification upgrading Average trainee age 31.972 18.257 24.551 26.26 28.449 (10.520) (2.430) (7.570) (10.360) (10.120) Average course duration (in hours) 40.413 809.58 1109.348 164.707 174.566 (44.970) (501.700) (508.450) (372.330) (201.130) Share of trainees working 0.689 0.271 0.550 0.464 0.491 (0.460) (0.440) (0.500) (0.500) (0.500) Average tenure of trainees (in months) 54.872 11.335 37.4 33.01 54.637 (70.560) (14.880) (46.290) (46.630) (54.630) Log hourly wage 2.545 1.683 2.341 2.119 2.238 (0.693) (0.449) (0.553) (0.567) (0.606) Share with schooling less than high school 0.237 0.442 0.149 0.344 0.259 (0.425) (0.497) (0.356) (0.475) (0.438) Share with schooling equal to high school 0.635 0.516 0.767 0.568 0.678 (0.481) (0.500) (0.423) (0.495) (0.467) Share with schooling more than high school 0.128 0.042 0.085 0.088 0.063 (0.334) (0.201) (0.279) (0.284) (0.243) Share employed in year before training 0.775 0.35 0.305 0.233 0.443 (0.417) (0.477) (0.455) (0.482) (0.479) Share employed in same firm in year before training 0.605 0.229 0.211 0.157 0.201 (0.488) (0.420) (0.499) (0.497) (0.456) Share employed in same occupation in year before training 0.513 0.133 0.125 0.074 0.157 (0.499) (0.340) (0.494) (0.482) (0.431) Share of workers in skill level 4 0.027 0.006 0.014 0.027 0.019 (0.162) (0.077) (0.116) (0.124) (0.112) Share of workers in skill level 3 0.118 0.062 0.192 0.086 0.086 (0.323) (0.241) (0.394) (0.274) (0.271) Share of workers in skill level 2 0.729 0.786 0.699 0.746 0.735 (0.444) (0.410) (0.458) (0.437) (0.441) Share of workers in skill level 1 0.125 0.145 0.094 0.169 0.178 (0.330) (0.352) (0.292) (0.366) (0.377) Number of trainees in manufacturing firms 2009 75,099 5,331 6,840 33,334 132,530 2010 166,694 7,660 9,358 49,237 46,670 2011 179,555 8,079 16,738 60,902 52,958 2012 164,784 7,550 11,568 50,279 61,214 Total 586,132 28,620 44,504 193,752 293,372 Age restriction No Yes Yes No No Education restriction No Yes Yes No Yes Note: Table reports means and standard devitations of attributes of SENAI trainees in manufacturing firms by training category. Skill levels 1 to 4 are based on the 1988 International Standard Classification of Occupations (see Appendix Table A1 for a detailed description). Table 3: Training activity in manufacturing firms by export status Non-exporters Exporters All firms Share of workers that received training by course category Technical upgrading 0.006 0.021 0.006 (0.039) (0.051) (0.040) Other categories 0.009 0.022 0.010 (0.050) (0.045) (0.050) Any category 0.015 0.043 0.016 (0.081) (0.093) (0.086) Share of workers that received training by occupation group Share of workers in skill level 4 0.006 0.011 0.007 (0.067) (0.068) (0.067) Share of workers in skill level 3 0.017 0.031 0.0197 (0.103) (0.099) (0.103) Share of workers in skill level 2 0.005 0.022 0.007 (0.044) (0.061) (0.046) Share of workers in skill level 1 0.005 0.018 0.006 (0.049) (0.076) (0.052) Observations 907,694 45,751 953,445 Notes: Standard deviations in parenthesis. Data refer to firms operating in the manufacturing sector over the period 2009-2012. A firm-year observation is classified as exporter if the firm had export revenue in that year. Skill levels 1 to 4 are based on the 1988 International Standard Classification of Occupations (see Appendix Table A1 for a detailed description). Table 4: Export participation and share of workers trained, OLS estimates for full sample (1) (2) (3) (4) A: Dependent variable: share of workers receiving technical upgrading Export (=1 if yes) 0.019*** 0.005*** 0.001** 0.001* (0.000) (0.000) (0.001) (0.001) Log employment 0.001*** 0.001*** (0.000) (0.000) B: Dependent variable: share of workers receving training in other categories Export (=1 if yes) 0.011*** 0.002*** -0.000 -0.000 (0.0003) (0.000) (0.001) (0.001) Log employment 0.001*** 0.000 (0.000) (0.000) Firm effects No No Yes Yes Sector effects No Yes No No State effects No Yes No No Firm controls No Yes No Yes Observations 953,445 953,445 953,445 953,445 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Firms controls are: average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; and share of workers in each of skill levels 1 to 4. Table 5: Export participation and share of workers trained, OLS estimates for switchers (1) (2) (3) (4) A: Dependent variable: share of workers receiving technical upgrading Export (=1 if yes) 0.002*** 0.001** 0.001** 0.001** (0.001) (0.000) (0.000) (0.000) Log employment 0.002*** 0.001 (0.000) (0.000) B: Dependent variable: share of workers receving training in other categories Export (=1 if yes) -0.000 -0.001 -0.001 -0.001 (0.001) (0.001) (0.001) (0.001) Log employment 0.001 0.002 (0.000) (0.001) Firm effects No No Yes Yes Sector effects No Yes No No State effects No Yes No No Firm controls No Yes No Yes Observations 29,347 29,347 29,347 29,347 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Firms controls are: average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; share of workers in each of skill levels 1 to 4. Table 6: Export participation and share of workers trained, IV estimates for switchers First stage (1) (2) Dependent variable: Export (=1 if yes) Industry-specific trade-weighted log real exchange rate 0.004*** 0.004*** (0.001) (0.001) F-statistic 11.312 11.171 Second stage A: Dependent variable: share of workers receiving technical upgrading Export (=1 if yes) 0.056* 0.058** (0.028) (0.029) B: Dependent variable: share of workers receving training in other categories Export (=1 if yes) 18.501 18.849 (19.465) (19.603) Firm effects Yes Yes Firm controls No Yes Observations 29,347 29,347 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; and share of workers in each of skill levels 1 to 4. Table 7: Alternative definition of dependent variable, IV estimates for switchers First stage (1) (2) Dependent variable: Export (=1 if yes) Industry-specific trade-weighted log real exchange rate 0.004*** 0.004*** (0.001) (0.001) F-statistic 11.312 11.171 Second stage A: Dependent variable: share of workers receiving technical upgrading Export (=1 if yes) 0.080** 0.083** (0.036) (0.037) B: Dependent variable: share of workers receving training in other categories Export (=1 if yes) 0.021 0.025 (0.027) (0.028) Firm effects Yes Yes Firm controls No Yes Observations 29,347 29,347 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; and share of workers in each of skill levels 1 to 4. Table 8: Export participation and share of workers trained by occupational group, IV estimates for switchers First stage (1) (2) Dependent variable: Export (=1 if yes) Industry-specific trade-weighted log real exchange rate 0.004** 0.004** (0.002) (0.002) F-statistic 10.673 10.771 Second stage A: Dependent variable: share of workers in skill level 4 receiving technical upgrading Export (=1 if yes) 0.052 0.049 (0.067) (0.065) B: Dependent variable: share of workers in skill level 3 receiving technical upgrading Export (=1 if yes) 0.086 0.097 (0.229) (0.241) C: Dependent variable: share of workers in skill level 2 receiving technical upgrading Export (=1 if yes) 0.061* 0.063* (0.034) (0.035) D: Dependent variable: share of workers in skill level 1 receiving technical upgrading Export (=1 if yes) 0.104** 0.106** (0.042) (0.042) Firm effects Yes Yes Firm controls No Yes Observations 18170 18170 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; and share of workers in each of skill levels 1 to 4. Table 9: Returns to technical upgrading within exporting firms, OLS estimates Dependent variable: log hourly wage (1) (2) (3) (4) (5) (6) Received technical upgrading (=1 if yes) 0.055*** 0.039*** 0.031*** 0.029*** 0.028*** 0.028*** (0.004) (0.005) (0.004) (0.005) (0.006) (0.006) Log employment 0.036*** 0.024*** 0.045*** (0.008) (0.002) (0.005) Worker controls No Yes No Yes No Yes Worker effects No No Yes Yes Yes Yes Firm controls No Yes No Yes No Yes Firm effects No No No No Yes Yes State effects No Yes No Yes No Yes Observations 14,009,134 14,009,134 14,009,134 14,009,134 14,009,134 14,009,134 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Worker controls are: age, race, gender, schooling level, and skill level based on occupation. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; share of workers in each of skill levels 1 to 4; and 2-digit industry dummies. Sample restricted to firms that exported in all years during 2009-2012. Table 10: Returns to technical upgrading within exporting firms by firm size, OLS estimates Dependent variable: log hourly wage (1) (2) (3) (4) (5) (6) A. Micro and small firms (employment < 100 workers) Received technical upgrading (=1 if yes) 0.070*** 0.048*** 0.054*** 0.053*** 0.054*** 0.053*** (0.007) (0.007) (0.011) (0.011) (0.012) (0.012) Observations 146,079 146,079 146,079 146,079 146,079 146,079 B. Medium size firms (employment between 100 and 500 workers) Received technical upgrading (=1 if yes) 0.060*** 0.050*** 0.044*** 0.044*** 0.044*** 0.043*** (0.005) (0.004) (0.006) (0.007) (0.007) (0.007) Observations 926,724 926,724 926,724 926,724 926,724 926,724 C. Large firms (employment > 500 workers) Received technical upgrading (=1 if yes) 0.053*** 0.037*** 0.029*** 0.028*** 0.027*** 0.027*** (0.004) (0.005) (0.005) (0.005) (0.006) (0.006) Observations 12,936,331 12,936,331 12,936,331 12,936,331 12,936,331 12,936,331 Worker controls No Yes No Yes No Yes Worker effects No No Yes Yes Yes Yes Firm controls No Yes No Yes No Yes Firm effects No No No No Yes Yes State effects No Yes No Yes No Yes Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Worker controls are: age, race, gender, schooling level, and skill level based on occupation. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; share of workers in each of skill levels 1 to 4; and 2-digit industry dummies. Sample restricted to firms that exported in all years during 2009-2012. Table A1: Description of ISCO skills based on occupations Skill Descritpion ISCO Major group Competence associated with general Skill level 1 education usually acquired by completion of (9) Elementary occupations compulsory education. (4) Clerks; (5) Service workers and shop and market sales workers; (6) Requires knowledge as for first skill level, but Skilled agriculture and fishery; (7) Skill level 2 typically a longer period of worker-related Craft and related workers; (8) Plant training or work experience. and machine operators and assemblers. Requires a body of knowledge associated (3) Technicians and associate Skill level 3 with a period of postcompulsory education professionals but not to degree level. Normally requires a degree or an equivalent (1) Legislators, senior officials and Skill level 4 period of relevant work experience. managers; (2) Professionals Table A2: Attributes of workers employed at exporting manufacturing firms by training status Did not receive Received Did not technical technical received any upgrading in upgrading in Full sample training in current or current or sample period previous years previous years Log hourly wage 2.130 2.136 2.603 2.154 (0.738) (0.734) (0.679) (0.737) Share with schooling less than high school 0.474 0.467 0.237 0.458 (0.499) (0.498) (0.425) (0.498) Share with schooling high school 0.415 0.423 0.632 0.430 (0.492) (0.494) (0.482) (0.495) Share with schooling more than high school 0.110 0.109 0.130 0.110 (0.313) (0.312) (0.336) (0.313) Share of production workers 0.458 0.458 0.427 0.457 (0.498) (0.498) (0.494) (0.498) Share of workers in skill level 4 0.046 0.045 0.030 0.044 (0.210) (0.207) (0.173) (0.206) Share of workers in skill level 3 0.069 0.070 0.129 0.072 (0.254) (0.256) (0.335) (0.260) Share of workers in skill level 2 0.736 0.736 0.737 0.736 (0.440) (0.440) (0.440) (0.440) Share of workers in skill level 1 0.147 0.147 0.102 0.145 (0.354) ().354 (0.303) (0.353) Tenure 53.329 53.073 65.031 53.516 (65.741) (65.420) (72.972) (65.754) Share of men 0.701 0.706 0.901 0.713 (0.457) (0.455) (0.298) (0.451) Share of white 0.643 0.645 0.669 0.646 (0.478) (0.478) (0.470) (0.478) Observations 11,439,308 13,489,458 519,676 14,009,134 Notes: Table reports means and standard deviations (in parentheses) for workers employed at manufacturing firms that exported every year over the period 2009-2012, by training status. Wages are in 2010 Brazilian Reais. Sample restricted to workers employed in firms that exported in all years over that period. Table A3: Returns to technical upgrading within exporting firms, OLS estimates with restricted sample (1) (2) (3) (4) (5) (6) Dependent variable: log hourly wage Received technical upgrading (=1 if yes) 0.056*** 0.040*** 0.033*** 0.031*** 0.029*** 0.029*** (0.004) (0.005) (0.005) (0.005) (0.005) (0.005) Log employment 0.036*** 0.023*** 0.046*** (0.008) (0.001) (0.000) Worker controls No Yes No Yes No Yes Worker effects No No Yes Yes Yes Yes Firm controls No Yes No Yes No Yes Firm effects No No No No Yes Yes State effects No Yes No Yes No Yes Observations 11,958,984 11,958,984 11,958,984 11,958,984 11,958,984 11,958,984 Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Worker controls are: age, race, gender, schooling level, and skill level based on occupation. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; share of workers in each of skill levels 1 to 4; and 2-digit industry dummies. Sample restricted to firms that exported in all years during 2009-2012. Table A4: Returns to technical upgrading within exporting firms by firm size categories, OLS estimates with restricted sample Dependent variable: log hourly wage (1) (2) (3) (4) (5) (6) A. Micro and small firms (employment < 100 workers) Received technical upgrading (=1 if yes) 0.072*** 0.048*** 0.056*** 0.056*** 0.056*** 0.055*** (0.008) (0.007) (0.011) (0.011) (0.011) (0.011) Observations 140,172 140,172 140,172 140,172 140,172 140,172 B. Medium size firms (employment between 100 and 500 workers) Received technical upgrading (=1 if yes) 0.062*** 0.051*** 0.047*** 0.046*** 0.046*** 0.045*** (0.005) (0.004) (0.006) (0.006) (0.006) (0.006) Observations 883,859 883,859 883,859 883,859 883,859 883,859 C. Large firms (employment > 500 workers) Received technical upgrading (=1 if yes) 0.054*** 0.038*** 0.031*** 0.030*** 0.029*** 0.028*** (0.004) (0.005) (0.005) (0.005) (0.006) (0.006) Observations 10,934,953 10,934,953 10,934,953 10,934,953 10,934,953 10,934,953 Worker controls No Yes No Yes No Yes Worker effects No No Yes Yes Yes Yes Firm controls No Yes No Yes No Yes Firm effects No No No No Yes Yes State effects No Yes No Yes No Yes Notes: Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions include year effects. Worker controls are: age, race, gender, schooling level, and skill level based on occupation. Firms controls are: log employment; average age of workers; average tenure; male to female workers ratio; share of workers with less than high school completed; share of workers with exactly high school completed; share of workers with more than high school completed; average log wage; share of white workers; share of workers in each of skill levels 1 to 4; and 2-digit industry dummies. Sample restricted to firms that exported in all years during 2009-2012. Figure 1: Real exchange rate, Brazil, 2009-2012 Panel A: Real effective exchange rate index, Jan 2009=100 105 100 95 90 85 80 75 70 65 Jul-09 Jul-10 Jul-11 Jul-12 Jan-09 Apr-09 Oct-09 Jan-10 Apr-10 Oct-10 Jan-11 Apr-11 Oct-11 Jan-12 Apr-12 Oct-12 Panel B: Selected bilateral real exchange rate indexes, Jan 2009=100 105 100 95 90 85 80 75 70 65 Jul-09 Jul-10 Jul-11 Jul-12 Jan-09 Apr-09 Oct-09 Jan-10 Apr-10 Oct-10 Jan-11 Apr-11 Oct-11 Jan-12 Apr-12 Oct-12 RMB EUR USD Notes: Indexes are based on data from the International Financial Statistics of the International Monetary Fund. A rise in the index denotes a depreciation of the real relative to foreign currencies.