Online Appendix Capital Account Opening and Wage Inequality Mauricio Larrain Columbia University October 2014 A.1 Additional summary statistics Tables 1 and 2 in the main text report summary statistics for the Chinn and Ito (2006) capital account openness index and aggregate wage inequality for each country, respectively. In this section, I report the evolution of both variables across countries and time. Table A.1 reports the evolution of the capital account openness index. According to the table, there is substantial variation of capital market openness both across countries and across time. Eastern European countries opened very quickly towards the end of the sample. Some countries (e.g., Denmark and Italy) opened in the 1980s. Other countries opened in the 1990s (e.g., Portugal and Spain). [Include Table A.1 here] Table A.2 reports the evolution of wage inequality across countries and time. Wage inequality between 1975 and 2005 increased in more than half of the countries in the sample. It increased particularly in Eastern European countries, but decreased in Scandinavian countries. The table also reports the evolution of wage inequality separately for manufacturing and non-manufacturing industries. For a given year, wage inequality in manufacturing sectors is on average higher than in non-manufacturing sectors. The large increase in wage inequality in Eastern European countries is heavily influenced by the increase in inequality in manufacturing industries. [Include Table A.2 here] A.1
A.2 Reverse causality In order to rule out concerns about reverse causality, I analyze whether pre-existing wage inequality in sectors with high financial dependence and strong complementarity explains the timing of the openings. In order to obtain a precise date of opening, I define the opening year as the year in which the Chinn and Ito (2006) openness index of acountryincreasesbymorethanonestandarddeviationacrossallcountriesandyears. Next, I rank all industries according to the product of both sectoral indices. Sectors with high financial dependence and strong complementarity are defined as those for which the product of the two sectoral indices is above the median of the product across sectors. Then, for each country I calculate average wage inequality in sectors with high financial dependence and strong complementarity before the opening year. Finally, I regress the capital opening year on pre-existing wage inequality in the aforementioned sectors. The results are reported in Table A.3. In column (1), I analyze the impact of the level of wage inequality on the opening year. The e ect is not statistically significant. In column (2), I analyze the impact of the change in wage inequality. Again, the e ect is not significant. According to these results, the timing of capital account opening does not vary with the degree of pre-existing wage inequality in sectors with high dependence and strong complementarity. [Include Table A.3 here] A.3 CSC estimation: system GMM In Section 3 of the main text, I estimated the skilled labor share equation (5) using adi erencegmmestimator. Inthissection,Ishowthatmyresultsarerobustto estimating the equation with a system GMM estimator, which combines the regression in di erences and in levels. For convenience, I re-write the skilled labor share equation here: ShareSkilled ct = + log(inequality) ct + log(capintensity) ct + c + t + " ct. For the system GMM estimation, the instruments for the regression in di erences are the same as in the text. The moment conditions for the regression in di erences are: E[z ct j " ct ]=0forj 2,t 3, where z =[ShareSkilled, Inequality, CapIntensity]. A.2
The instruments for the regression in levels are the lagged di erences of the corresponding variables. These are appropriate instruments under the additional assumption of no correlation between the di erences of these variables and the country fixed e ects. This leads to the following additional moment conditions for the regression in levels: E[ z ct j " ct ]=0forj 2,t 3. Table A.4 reports the estimates of the skilled labor share equation for each industry. The coe cient of correlation between this alternative capital-skill complementarity index and the original index is 0.92. In Table A.5, I re-estimate the e ect of capital account opening on sectoral wage inequality, using the alternative complementarity measure. According to the results, the coe cient of the triple interaction term is significant and very similar in magnitude than the coe cient obtained using the original complementarity measure (see Table 8). Therefore, my results are robust to estimating the complementarity index with system GMM. [Include Tables A.4 and A.5 here] A.4 Compositional e ects The transmission mechanism that I highlight in this paper works at the firm level. An alternative story is that capital opening induces skill-intensive firms to expand. This would change the within-industry composition of firms towards more skill-intensive firms, increasing the relative demand for skilled labor. This compositional e ect could lead to higher wage inequality, even if firms do not exhibit capital-skill complementarity. As before, there would be a threat to identification only if this between-firm e ect is particularly strong in industries with high external dependence and strong complementarity. To disentangle the within-firm and between-firm channels, I would have to conduct the analysis at the firm level. Unfortunately, I do not have access to a cross-country firm-level dataset with wage inequality information. However, I do have access to afirm-leveldatasetforanemergingmarket,chile,whichcontainswageinformation for production (i.e., blue-collar) and non-production (i.e., white-collar) workers. The dataset, called ENIA, is an annual survey covering all manufacturing firms in Chile. 1 1 ENIA stands for Encuesta Nacional Industrial Anual. The data covers the period 1979-1998 and comes from Alvarez and Crespi (2007). A.3
Iconductafirm-levelestimationtoanalyzethee ectofopeningonwithin-firmoutcomes: log(x) jit = Openness t Z i + j + t + " jit, where x is either capital stock per unit of skilled labor or wage inequality of firm j in industry i in year t. 2 Z i denotes either external financial dependence or capital-skill complementarity of industry i. Thespecificationincludesasetoffirmfixede ects( j ) and year fixed e ects ( t ). Firm fixed e ects control for all time-invariant company characteristics. I cluster standard errors at the industry level. Table A.6 reports the results. In column (1), I exploit variation in external financial dependence across sectors and analyze the e ect on the capital stock. I find that capital opening increases within-firm capital stock particularly in externally dependent sectors. In column (2), I exploit cross-sectoral variation in complementarity and estimate the e ect on wage inequality. 3 According to the results, opening increases within-firm wage inequality particularly in sectors with strong complementarity. These findings support the within-firm channel. [Include Table A.6 here] Finally, I analyze whether opening the capital account leads skill-intensive firms to expand. To measure skill intensity, I use the lagged ratio of skilled to unskilled labor. I estimate the e ect on firm-level production: log(output) jit = Openness t SkillIntensity j,t 1 + j + t + " jit Column (3) of Table A.6 reports the results. The results show that skill-intensive firms do not expand more than non skill-intensive firms after capital opening. This finding provides evidence against the between-firm composition channel. A.5 E ect across demographic groups Capital opening a ects wage inequality di erentially across industries only if labor is imperfectly mobile across sectors. In the economy, there are some workers that are 2 I define wage inequality as the relative wage between non-production and production workers. 3 Since there is not su cient cross-sectional variation in the external dependence and complementarity indices within manufacturing, I only exploit variation in complementarity across sectors. A.4
more mobile than others. As a result, the di erential e ect on relative wages should vary across di erent types of workers. I focus on two demographic characteristics: age and sex. Since young workers have accumulated less sector-specific human capital than older workers, they are more mobile across sectors. Regarding sex, existing evidence shows that men tend to be more mobile than women. I divide each country-sector-year cell in the sample according to these demographic characteristics and re-estimate Equation (11) for each group. Column (1) of Table A.7 reports the results for young workers, which are defined as workers under the age of 50. Column (2) reports the results for older workers, those above 50. According to the results, the e ect is large and significant for older workers, but is not significant for young workers. Next, I divide the sample by sex. In column (3), I show the results for male workers and in column (4) for female workers. Although the e ect is significant for both groups, the magnitude of the e ect for women is more than twice the magnitude of the e ect for men. [Include Table A.7 here] A.6 Alternative samples In Table A.8, I re-estimate the sectoral regressions excluding the most developed countries from the sample. I rank all countries according to PPP-based GDP per capita in 1990 (the mid-year of the sample). In column (1), I drop the most developed country from the sample. In columns (2), (3), (4), I drop the two, three, and five most developed countries. According to the World Bank s WDI data, the countries with highest PPP-based GDP per capita in 1990 are (in order): Denmark, Belgium, Austria, Italy, and Japan. The magnitude of the triple interaction terms remains roughly constant across all four alternative samples and the e ect is always highly significant. This indicates that the results are driven by the less-developed, capital-scarce economies of the sample. [Include Table A.8 here] A.5
Table A.1: Capital Account Openness Index by Country and Time (1) (2) (3) (4) (5) (6) Average Average Average Average Cumulative Cum. Change 1970s 1980s 1990s 2000s Change (Annualized) Australia -0.106 1.042 2.191 1.132 1.239 0.041 Austria 1.132 1.132 2.059 2.456 1.323 0.044 Belgium 1.016 0.521 2.218 2.191 0.696 0.023 Czech Republic.. -0.106 1.808 2.562 0.285 Denmark -0.106 0.221 2.376 2.456 2.562 0.085 Finland 0.885 1.132 2.059 2.456 2.562 0.085 France -0.106-0.212 1.819 2.456 2.562 0.085 Greece -1.159-1.159 0.022 2.105 3.615 0.121 Hungary. -1.856-0.947 1.588 4.311 0.227 Ireland -0.524-0.106 1.555 2.456 3.258 0.109 Italy -1.856-0.456 1.819 2.456 4.311 0.144 Japan 1.185 2.297 2.323 2.456 1.323 0.044 Korea -0.738-0.738-0.528-0.106 1.053 0.035 Poland. -1.856-1.297-0.334 1.935 0.102 Portugal -1.159-0.949 1.422 2.456 3.615 0.121 Slovakia.. -1.159-0.275 2.033 0.226 Slovenia.. -0.370 1.397 3.086 0.343 Spain -0.528-0.106 1.290 2.456 3.615 0.121 Sweden 1.132 1.132 1.794 2.456 1.323 0.044 United Kingdom -0.363 2.297 2.456 2.456 3.258 0.109 Notes: the table reports the evolution of the capital account openness index for the 20 countries in the sample during the period 1975-2005. The openness index comes from Chinn and Ito (2006) index. Columns (1)-(4) report the average of the index over di erent decades; Column (5) reports the cumulative change of the index; Column (6) reports the annualized cumulative change. A.6
Table A.2: Aggregate Wage Inequality by Country and Time (1) (2) (3) (4) (5) (6) (7) Average Average Average Average Cum. Change Cum. Change Cum. Change 1970s 1980s 1990s 2000s All Sectors Manuf. Non-manuf. Australia. 1.463 1.474 1.551 0.006-0.001 0.010 Austria. 1.650 1.557 1.488-0.005-0.004-0.007 Belgium. 1.478 1.528 1.499 0.001 0.006-0.000 Czech Republic.. 2.239 2.283 0.007 0.010 0.004 Denmark. 1.581 1.448 1.415-0.009-0.009-0.009 Finland 1.870 1.697 1.550 1.565-0.010-0.004-0.014 France. 1.882 1.865 1.767-0.013-0.014-0.017 Greece.. 1.570 1.527-0.001-0.007-0.009 Hungary.. 2.315 2.477 0.030 0.004 0.009 Ireland. 1.716 1.825 1.864 0.012 0.006 0.009 Italy 1.255 1.197 1.235 1.429 0.007-0.010-0.005 Japan 1.694 1.660 1.701 1.628-0.003-0.001-0.002 Korea 2.014 1.962 1.632 1.526-0.013-0.009-0.016 Poland.. 1.573 1.546 0.001 0.009 0.006 Portugal.. 2.243 2.392 0.023-0.010-0.009 Slovakia.. 1.796 1.776 0.004 0.013 0.003 Slovenia.. 2.102 2.134-0.010-0.034-0.013 Spain. 1.435 1.615 1.670 0.014 0.004 0.005 Sweden. 1.538 1.543 1.465-0.003-0.007-0.009 United Kingdom 1.836 1.861 1.834 1.950 0.001-0.002 0.004 Notes: thetablereportstheevolutionofaggregatewageinequalityforthe20countriesinthesampleduringtheperiod1975-2005.wageinequalityisdefined as the relative wage between workers with college and high-school education. Columns (1)-(4) report average wage inequality over di erent decades; Column (5) reports the annualized cumulative change over the entire sample period for all sectors; Column (6) reports the change for manufacturing sectors; Column (7) reports the change for non-manufacturing sectors. A.7
Table A.3: E ect of Pre-existing Wage Inequality on Timing of Capital Account Opening (1) (2) Pre-existing Inequality Level 0.0014 (0.0061) Pre-existing Inequality Change 0.0003 (0.0009) Observations 20 20 R-squared 0.199 0.070 Notes: the table reports the estimates of the e ect of pre-existing wage inequality in sectors with high financial dependence and strong complementarity on the year of capital account opening. Sectors with high financial dependence and strong complementarity are defined as those for which the product of the two sectoral indices is above the median of the product across sectors. The opening year is defined as the year in which the Chinn and Ito (2006) openness index of a country increases by more than one standard deviation of the index across all countries and years. Column (1) uses as independent variable the pre-existing level of wage inequality; Column (2) uses as independent variable the pre-existing change in wage inequality. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.8
Table A.4: Sectoral Index of Capital-skill Complementarity: Calculated Using System GMM (1) (2) (3) (4) (5) (6) Industry Industry Coe. StdDev. Coe. StdDev. Number of Name ISIC code Beta Beta Gamma Gamma Observations (CSC index) Manufacturing of wood 20 0.425*** 0.057 0.143*** 0.025 317 Manufacturing of coke, refined petroleum 23 0.87*** 0.081 0.022* 0.012 317 Manufacturing of chemicals 24 0.901*** 0.063 0.065** 0.03 317 Manufacturing of rubber, plastics 25 0.729*** 0.074 0.3*** 0.029 317 Manufacturing of other non-metallic mineral prod. 26 0.58*** 0.086 0.249*** 0.031 317 Manufacturing of machinery and equipment 29 0.782*** 0.059-0.106*** 0.017 317 Construction 45 1.542*** 0.097 0.08** 0.032 317 Wholesale trade and commission trade 51 1.725*** 0.066 0.171*** 0.03 317 Retail trade, except of motor vehicles 52 1.958*** 0.056-0.187*** 0.026 317 Hotels and restaurants 55 0.939*** 0.046 0.16*** 0.027 317 Post and telecommunications 64 0.977*** 0.066 0.345*** 0.031 317 Real estate activities 70 1.357*** 0.028-0.013 0.023 317 Education 80 0.692*** 0.033-0.077*** 0.011 317 Health and social work 85 0.638*** 0.054-0.096*** 0.019 317 Notes: the table reports the estimates of the skilled labor share equation for each of the 15 two-digit industries in the sample. The estimation is conducted using system GMM. Columns (2) and (3) report the coe cient and standard deviation of the elasticity of the share of wages paid to skilled labor with respect to relative wages; Columns (4) and (5) report the coe cient and standard deviation of the elasticity with respect to capital intensity. The coe cient of column (4) corresponds to the capital-skill complementarity index. Standard errors are clustered at the country level. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.9
Table A.5: E ect of Capital Account Opening on Sectoral Wage Inequality: Complementarity Index Calculated Using System GMM (1) (2) (3) (4) Capital Openness 0.007 0.006 (0.010) (0.010) Fin. Dep. -0.099*** (0.012) Comp. 0.074*** (0.027) Capital Openness * Fin. Dep. -0.033-0.014-0.033-0.014*** (0.019) (0.020) (0.020) (0.004) Capital Openness * Comp. 0.003 0.011 0.003 0.011 (0.021) (0.025) (0.022) (0.010) Capital Openness * Fin. Dep. * Comp. 0.201*** 0.120 0.201*** 0.120*** (0.060) (0.071) (0.062) (0.029) Fixed E ects Country Yes No No No Year Yes No No No Country-year No No Yes Yes Country-industry No Yes Yes Yes Industry-year No Yes No Yes Observations 6,480 6,480 6,480 6,480 R-squared 0.899 0.898 0.932 0.933 Notes: thetablereportstheestimatesofthee ectofcapitalaccountopeningonsectoralwageinequality.thecapitalskill complementarity index is calculated using system GMM. Column (1) includes country and year fixed e ects; Column (2) includes country-industry and industry-year fixed e ects; Column (3) includes country-year and countryindustry fixed e ects; Column (4) includes country-year, country-industry, and sector-year fixed e ects. Fin. Dep. stands for external financial dependence and Comp. for capital-skill complementarity. Standard errors in parentheses are clustered at the country level. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.10
Table A.6: E ect of Capital Account Opening on Sectoral Wage Inequality: Firm-level Evidence (1) (2) (3) Capital Wage Output Stock Inequality Capital Openness * Fin. Dep. 0.048* (0.028) Capital Openness * Comp. 0.119** (0.059) Capital Openness * Skill intensity(-1) 0.002 (0.002) Fixed E ects Firm Yes Yes Yes Year Yes Yes Yes Observations 14,217 13,321 11,692 R-squared 0.850 0.627 0.954 Notes: the table reports the estimates of the e ect of capital account opening on firm-level outcomes in Chile. The firm-level data comes from ENIA. Columns (1), (2), (3) estimate the e ect on the capital stock per unit of skilled labor, wage inequality, and production, respectively. Fin. Dep. stands for external financial dependence; Comp. stands for capital-skill complementarity; Skill intensity(-1) stands for lagged ratio of skilled to unskilled labor. All specifications include firm and year fixed e ects. Standard errors in parentheses are clustered at the firm level. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.11
Table A.7: E ect of Capital Account Opening on Sectoral Wage Inequality: Demographic Groups (1) (2) (3) (4) E ect by age E ect by sex Young Old Male Female Workers Workers Workers Workers Capital Openness * Fin. Dep. 0.020-0.038-0.040-0.151** (0.024) (0.036) (0.029) (0.070) Capital Openness * Comp. 0.099** -0.001 0.022-0.224** (0.036) (0.019) (0.035) (0.097) Capital Openness * Fin. Dep. * Comp. -0.059 0.219* 0.213* 0.612** (0.113) (0.118) (0.099) (0.245) Fixed E ects Country-year Yes Yes Yes Yes Country-industry Yes Yes Yes Yes Industry-year Yes Yes Yes Yes Observations 5,210 5,223 5,234 5,223 R-squared 0.923 0.933 0.936 0.971 Notes: thetablereportstheestimatesofthee ectofcapitalaccountopeningonsectoralwageinequalityfordi erent demographic groups. Columns (1)-(2) divide the sample by age; Columns (3)-(4) by sex. Column (1) includes workers younger than 50; Column (2) workers older than 50. Column (3) includes male workers; Column (4) female workers. Fin. Dep. stands for external financial dependence and Comp. stands for capital-skill complementarity. All specifications include country-year, country-industry, and industry-year fixed e ects. Standard errors in parentheses are clustered at the country level. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.12
Table A.8: E ect of Capital Account Opening on Sectoral Wage Inequality: Alternative Samples (1) (2) (3) (4) Excluding Excluding Excluding Excluding most two most three most five most developed developed developed developed country countries countries countries Capital Openness * Fin. Dep. -0.026** -0.028** -0.028** 0.001 (0.008) (0.008) (0.008) (0.007) Capital Openness * Comp. 0.037** 0.044** 0.043** 0.023* (0.010) (0.010) (0.010) (0.013) Capital Openness * Fin. Dep.*Comp. 0.109*** 0.125*** 0.121*** 0.094*** (0.027) (0.027) (0.027) (0.023) Fixed e ects Country-year Yes Yes Yes Yes Country-industry Yes Yes Yes Yes Industry-year Yes Yes Yes Yes Observations 6,090 5,700 5,310 3,615 R-squared 0.959 0.958 0.959 0.963 Notes: thetablereportstheestimatesofthee ectofcapitalaccountopeningonsectoralwageinequalityfordi erent sample of countries. Column (1) excludes the most developed country from the sample (highest PPP-based GDP per capita in 1990, the mid-year of the sample); Column (2) excludes the two most developed countries; Column (3) excludes the three most developed countries; Column (4) excludes the five most developed countries. Standard errors in parentheses are clustered at the country level. ***, **, * denote statistical significance at 1%, 5%, and 10%. A.13