Policy Research Working Paper 10332 Trade and Local Labor Market Outcomes in Mexico Disentangling the Channels and the Role of Geography, Sectors, and Trade Types Emmanuel Vazquez Deborah Winkler Macroeconomics, Trade and Investment Global Practice February 2023 Policy Research Working Paper 10332 Abstract This study provides new evidence on the local labor market immigration and lowered the rate of informal workers. impacts of trade, differentiating between the employment, Second, the analysis examines differences by geography and income, migration, and informality channels. It uses a sectors. It finds that trade affected labor markets in the unique dataset matching information on exports and North through the income and migration channels and in imports from customs with indicators on employment and the South through the employment and informality chan- labor incomes for around 2,000 Mexican municipalities nels. Exports benefitted the total incomes of workers in over 2004–14. The analysis uses an instrumental vari- both the manufacturing and service sectors but reduced able approach that combines the initial structure of trade informality only in manufacturing. Third, the study sug- across municipalities with global trends in trade between gests a more favorable role of intermediate relative to final low- and middle-income countries (excluding Mexico) imports, driven by manufacturing imports. It also finds and the United States by sector. First, the study finds that evidence for positive spillovers from global value chain par- expanding exports per worker in Mexico’s municipalities ticipation through the employment and income channels. increased labor force participation but not employment Finally, it examines how local policy mediates the labor rates. Exports also raised total labor incomes but not aver- market effects from trade, focusing on connectivity, labor age labor incomes, implying a growing labor supply. The market flexibility, and education spending. results also find that export and import expansion increased This paper is a product of the Macroeconomics, Trade and Investment Global Practice. 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://www.worldbank.org/prwp. The authors may be contacted at evazquez@cedlas.org and dwinkler2@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 Trade and Local Labor Market Outcomes in Mexico: Disentangling the Channels and the Role of Geography, Sectors, and Trade Types Emmanuel Vazquez1 and Deborah Winkler2 Keywords: International trade, exports, imports, global value chains, local impacts, labor market, employment, income, migration, informality, flexibility, education spending, connectivity, Mexico. JEL Codes: F14, F16, D3, I3, J61 1 Senior Researcher, Center for Distributive, Labor, and Social Studies, Instituto de Investigaciones Económicas, Facultad de Ciencias Económicas, Universidad Nacional de La Plata. E-mail: evazquez@cedlas.org. 2 Senior Economist, Global Trade and Regional Integration, World Bank Group. Email: dwinkler2@worldbank.org. The authors would like to thank Carlos Rodríguez-Castelán and Hernan Winkler for their permission to build on and expand their research in Rodríguez-Castelán et al. (2020). They also thank Luis Alejandro Aguilar Luna for his excellent research assistance in computing trade and global value chain measures drawing on Economic Census data. The authors are also grateful to Claudia Berg, Raymundo M. Campos-Vasquez, Sébastien Dessus, Christoph Lakner, Gladys C. Lopez-Acevedo, Maryla Maliszewska, Ambar Narayan, Israel Osorio-Rodarte, Raymond Robertson and Ben Shepherd for their helpful comments and suggestions, as well as to the participants of the World Bank authors’ workshops “Leveraging trade for more and better job opportunities in developing countries” in May and November 2022. They thank Natalia Volkow and Liliana Martinez of the Microdata Center of the National Institute of Statistics and Geography, Mexico, for their support for data use in compliance with the confidentiality requirements set by Mexican law. The work was carried out under the overall supervision of Sébastien Dessus and Antonio Nucifora and has been cleared by Mark Thomas (Country Director of Mexico). The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank Group and its affiliated organizations, its Executive Directors or the governments they represent. 1. Motivation The empirical literature is inconclusive about the labor market effects of trade. Trade affects labor markets through relative price changes, but the net effect depends on the magnitude of the underlying channels, and which sectors and locations are expanding or contracting (Caliendo, Dvorkin and Parro 2019). Contrasting findings can also be attributed to the different types of trade and levels of development under study. Evidence for low- and middle-income countries often suggests a positive effect of exports and trade openness for employment, at least at the aggregate level and over the long term (Dutt et al. 2009; Hollweg et al. 2014). In contrast, recent empirical evidence finds no impact of participation in global value chains (GVCs) on employment in low- and middle-income countries (Pahl and Timmer 2020) or highlights the negative effects of import competition from low- and middle-income countries on workers in developed economies (e.g., Autor, Dorn, and Hanson 2013; Balsvik, Jensen, and Salvanes 2015; Hakobyan and McLaren 2016). Similarly, other studies find that trade liberalization could lead to heightened inequality and deteriorating labor market outcomes (De Loecker et al. 2016; Dix-Carneiro and Kovak 2017; Erten, Leight and Tregenna 2019; Goldberg and Pavcnik 2007; Nicita 2009; Pavcnik 2017; Revenga 1997; Topalova 2010). Another explanation for diverging effects could be the focus on labor demand versus labor supply. Trade expansion not only affects labor demand, but also the supply of jobs and the geography of workers within countries. Labor reallocations across sectors and regions as well as changes in unemployment rates are reflected in models at the sub-national level, following the Autor, Dorn and Hanson (2013) methodology. Our paper contributes to this vibrant literature by exploiting variations in trade growth across almost 2,000 Mexican municipalities over the period 2004-2014 to identify different local labor market impacts. Building on and expanding the study by Rodríguez-Castelán et al. (2020), this study contributes to a better understanding of the underlying channels through which higher trade integration affects labor markets in low- and middle-income countries, differentiating between (i) employment, (ii) labor incomes, (iii) migration, and (iv) informality. It applies an instrumental variable (IV) approach that combines the initial structure of trade across municipalities with global trends in sectoral trade between low- and middle-income countries (excluding Mexico) and the United States. A key finding is that trade in a middle-income country setting affects local labor markets more strongly through increased inward migration and movements out of the informal to the formal labor market, rather than through higher employment or labor incomes. Mexico is ideally suited to examine the labor market effects of trade in a middle-income country, given its strong trade openness and integration in GVCs. Mexico entered the General Agreement on Tariffs and Trade (GATT) in 1986 and the North American Free Trade Agreement (NAFTA) in 1994, resulting in significant trade liberalization, increased trade openness, and export diversification away from oil. Mexico’s goods and services export and import share in gross domestic product (GDP) expanded by a staggering 20 percentage points since 1990 (Figure 1), with more than three-quarters of trade depending on the United States. However, Mexico’s labor market outcomes are puzzling. First, Mexico has underperformed other countries in Latin America and the Caribbean (LAC) in terms of poverty reduction (World Bank 2019).3 If the positive trade-labor market nexus holds, one would expect larger declines in poverty for more trade-integrated countries. Second, labor supply has expanded, as evidenced by the increase in Mexico’s labor force participation rate (% of population ages 15-64) from 60.6 to 65.25 percent 3 Between 2004 and 2014, poverty in Mexico declined by 4.0 percentage points, from 37.6 percent to 33.6 percent, compared to nearly 17.0 percentage points in LAC, from 41.3 percent to 24.4 percent (applying the US$5.50-a-day per capita poverty line in 2011 purchasing power parity). Moreover, almost half the decline in poverty in Mexico is explained by redistribution (because of the shift from general subsidies to targeted and conditional transfers) rather than to economic growth, despite the expansion in trade over the period. In Latin America, in contrast, redistribution only explains about 20.0 percent of the decline, while economic growth accounted for nearly 80.0 percent of the reduction in poverty. 1 over the period 1990-2019. At the same time, labor demand related to integration in global trade networks is geographically concentrated in the North, close to the U.S. border. Figure 1: Goods and services exports and imports in Mexico, 1990-2018 Exports Imports 45 600 45 600 40 40 500 35 35 30 400 30 400 25 25 300 20 20 15 200 15 200 10 10 100 5 5 0 0 0 0 2000 1990 1992 1994 1996 1998 2002 2004 2006 2008 2010 2012 2014 2016 2018 2014 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2016 2018Exports (% of GDP) Exports (billions 2015 usd) Imports (% of GDP) Imports (billions of 2015 usd) Source: Own calculations based on World Development Indicators. This paper contributes to the literature in several ways. First, it provides a comprehensive framework and new insights into the underlying channels through which trade affects labor markets and broader development, distinguishing between the employment, labor income, migration and informality channels. Our study finds that expanding exports per worker in Mexico’s municipalities increased labor force participation, but not employment rates. Exports also raised total labor incomes, but not average labor incomes, implying a growing labor supply. In contrast, we find no effect of growing imports per worker through the employment and income channels. We therefore test whether growing labor supply through inward migration and movements from the informal to the formal labor market dilute the impact of trade through the employment and income channels. Our findings suggest that both export and import expansion per worker increased immigration to Mexican municipalities, driven by domestic rather than international unskilled migration. These findings support the hypothesis that the lack of positive labor market impacts of trade integration may be partly explained by migration (Robertson 2007) and complement the results by Majlesi and Narciso (2018) who find that Mexican municipalities that are more strongly exposed to Chinese import competition experience higher emigration to other municipalities, but less emigration to the United States. We also find that higher exports and imports per worker decreased the rate of informal workers, confirming studies for South Asian countries that find declines in informality, but no increases in employment due to export expansion (Artuc et al. 2019, Robertson et al. 2020). Second, it contributes to the number of studies assessing the impact of trade on labor markets for Mexico specifically. Lederman, Maloney, and Servén (2005) argue that NAFTA had positive impacts on foreign direct investment (FDI), trade, and productivity in Mexico, but only modest impacts on wages. Industry-level evidence points to a small but positive impact of FDI on manufacturing employment (Waldkirch et al. 2009). Firm or plant level studies find a connection between trade and increased wage inequality (Frías et al. 2012; Verhoogen 2008). Other studies on Mexico focused on the links between globalization and labor markets at the regional level. Chiquiar (2008) finds that regions exposed to NAFTA show an increase in wage levels, but a decrease in the skill premium. Airola (2008) finds that the wage skill premium is negatively related to Maquiladora employment (used as a proxy for FDI). The latter findings are consistent with the Stolper-Samuelson theorem of the Heckscher-Ohlin model, suggesting that factor prices of factors used more intensively in production increase, which in Mexico’s case is lower- 2 skilled labor. In contrast, Feenstra and Hanson (1997) find a significant link between higher FDI and the expansion of skilled labor in total wages. None of these studies, however, identifies the causal impact of the growth in exports and imports experienced by Mexican municipalities on local labor markets. Exceptions include newer studies which examine the local labor market outcomes in Mexico following growing import competition from China in the United States (Mendez 2015, Chiquiar, Covarrubias, and Salcedo 2017, and Robertson, Halliday and Vasireddy 2019). Third, this paper adds to the scarce number of studies on low- and middle-income countries investigating the effect of trade on local labor market outcomes. Recent studies for China and Vietnam find a positive effect of exports on labor market outcomes and poverty reduction at the subnational level (Erten and Leight 2019; McCaig 2011, McCaig and Pavcnik 2018) and on reduced poverty and inequality at the state-level in Brazil (Castilho, Menéndez, and Sztulman 2012). Other recent papers find that increasing exports per worker result in higher wages for workers generally and in declining informality in India and Sri Lanka (Artuc et al. 2019) and Bangladesh (Robertson et al. 2020). In contrast, a recent study on the Arab Republic of Egypt finds no significant effect of rising exports on wages, informality, and female labor force participation at the local level (Robertson et al. 2021). Fourth, the analysis also sheds light on differences by geography and sectors. The study finds that trade affected labor markets in the North through the income and migration channels, while it affected the South through the employment and informality channels. Specifically, total labor incomes and the number of immigrants in Northern municipalities expanded due to both higher exports and imports per worker. In contrast, export expansion in the South benefited labor force participation and employment rates (and to a lesser extent also migration), while increasing imports per worker hurt them. In a second stage, we assess whether a municipality’s trade expansion affects labor market outcomes differently across sectors. Our findings suggest that export expansion in Mexico benefited total labor incomes in both the manufacturing and service sectors. Since the trade data in the dataset mainly consist of goods, this finding could point to an indirect linkage effect, i.e., goods exports may rely on domestic services inputs, thereby benefitting total labor incomes in service sectors. Expanding exports per worker also reduced the informality rate in the manufacturing sector. Fifth, this study not only examines the impact of exports but also of imports, especially of different types of imports and GVC spillovers more broadly. The absence of effects of imports through the employment and income channels could point to a more nuanced role of imports. The latter include not only final imports that can compete with domestic production, but also imported inputs used for domestic or export production. To explore this question further, we distinguish between intermediate and final imports drawing on the Economic Census data. The direction of coefficients suggests a more favorable role of intermediate relative to final imports through the employment channel, driven by manufacturing imports. In contrast, we find a positive and significant role of both intermediate and final services imports through the employment channel. In addition, the analysis uses GVC participation measures drawing on Economic Census data at the state level and finds evidence for positive GVC spillovers on labor markets at the municipality level through the employment channel. This complements evidence at the firm-level suggesting that GVC-participating firms employ more workers relative to exporter-only, importer-only and purely domestic firms (World Bank 2020). We also find a positive association with total labor incomes, but a negative one with average labor incomes, which again may be linked to growing labor supply through the migration channel. Finally, this study explores the role of selected policy factors that mediate the impact of trade on local labor markets. We focus on three policy areas, namely connectivity, labor market flexibility and spending on education. Connectivity has been identified as a determinant of trade and GVC participation across a large set of countries (Fernandes, Kee and Winkler 2021), so workers in locations with better connectivity may benefit more strongly from trade expansion. We find that higher imports per worker 3 increase total labor incomes and the number of immigrants more strongly in municipalities with higher initial road networks, which could be linked to imports for export production (GVC participation), as lead firms require good connectivity to be able to deliver their output in a timely and reliable manner. Labor market flexibility determines how quickly firms and workers react to trade shocks in terms of adjustments in the quantity of workers and is thus expected to lead to different labor market outcomes. In line with expectations, results imply that the quantity of labor (measured by migration and the informality rate) reacts more strongly to export shocks in municipalities with high labor market flexibility, while the price of labor (measured by average labor incomes) reacts more strongly in municipalities with low labor market flexibility. Spending on education can shape the quality of labor supply in a region which, in turn, is a driver of trade and GVC participation (Fernandes, Kee and Winkler 2021). It thus also mediates the labor market effects from trade. Our findings suggest that export expansion benefits regions with higher spending on education through increased total labor incomes, while regions spending less on education gain from higher employment and reduced informality rates. The migration channel is stronger in regions with lower spending on education, implying that the availability of skills in a region acts as a deterrent for immigrants who are predominantly unskilled. The rest of this paper is structured as follows. Section 2 presents descriptive statistics, the methodology and the data used. Section 3 describes the local labor market effects of both expanding exports and imports per worker along the employment, income, migration, and informality channels. Section 4 presents extensions allowing for variation along the dimensions of geography (North-South), sectors of employment, labor incomes and informality, as well as types of imports (final versus intermediates) and GVC spillovers. It also assesses the role of selected policy measures in mediating the trade-labor relationship, focusing on connectivity, labor market flexibility, and spending on education. Section 5 concludes and discusses policy implications. 2. Descriptive statistics, methodology and data 2.1 Descriptive statistics This section differentiates between four channels – employment, income, migration and informality – to assess the possible role of trade descriptively. The assessment compares labor market outcomes in municipalities with high exports per worker (i.e., above the median) with those characterized by low exports per worker (i.e., below the median). Focusing on the employment channel, Figure 2 shows that the average labor force participation and employment rates in municipalities with high exports per workers is 5 percentage points higher than in municipalities with low exports per workers (left panel). The difference is 8 percentage points for imports (right panel). A visual inspection of observations of exports per worker across all Mexican municipalities and their labor force participation and employment rates also suggests a positive correlation (Appendix 1, left panel). The association is positive, but weaker for imports per worker (right panel). Focusing on the income channel in a second step, the difference in total labor incomes for municipalities with high trade per worker relative to those with low trade per worker is striking, both for exports and imports (Figure 3, upper panel). We can also detect a difference in average labor incomes with high versus low trade per worker, although the difference is larger for imports than for exports (bottom panel). The difference is 22 percent for exports, and 40 percent for imports. The scatterplot across all Mexican municipalities suggests a positive correlation between exports per worker and total labor incomes. In contrast, there seems to be no relationship with average labor incomes (Appendix 2, 4 left panel). The association also seems to be non-existent between total and average labor income and imports per worker (right panel). Figure 2: Exports, imports and average job outcomes, LFP and employment rate Exports Imports Labor force participation Labor force participation 50 45.5 50 46.8 40.2 38.9 40 40 30 30 High exports High imports 20 Low exports 20 Low imports 10 10 0 0 Employment rate Employment rate 50 50 44.5 43.3 38.1 36.8 40 40 30 High exports 30 High imports 20 Low exports 20 Low imports 10 10 0 0 Source: Own elaboration based on Population Census, Integrated Public Use Microdata Series and Customs data. Notes: (1) The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period. (2) A municipality has relatively high exports (imports) if its exports (imports) per 2000 worker ratio in 2015 is above the median across municipalities and low exports (imports) if its ratio lies below the median. Third, a municipality’s trade integration could strongly matter for migration patterns, both across domestic municipalities and with other countries, specifically the United States. More trade opportunities in a municipality may deter migration outflows, while attracting more migrants from other countries or pulling back Mexican migrants from the United States. In fact, the number of unauthorized migrants from Mexico to the United States declined from 6.9 million in 2007 to 5.6 million by 2015 (Gonzalez-Barrera and Krogstad 2019). Trade in a municipality may also have effects on remittances and the working age population. We therefore investigate the migration channel and again detect a clear difference between municipalities that are highly integrated in trade and those that are not. The average number of immigrants from other Mexican states in municipalities with high exports per worker is 142, while that in municipalities with low exports per worker is less than half this magnitude, namely 63 (Figure 4, upper left panel). Again, the relative difference is larger for imports (upper right panel). The analysis also shows a difference between the number of immigrants from abroad, which is less pronounced (bottom panel) and visually confirmed in the scatterplots across all municipalities (Appendix 3). 5 Figure 3: Exports, imports and average job incomes Exports Imports Total labor income Total labor income 15,000,000 15,000,000 10,780,209 12,001,126 10,000,000 10,000,000 High exports High imports 5,000,000 4,099,168 Low exports Low imports 5,000,000 2,878,250 0 0 Average labor income Average labor income 6,000 6,000 5,196 4,888 5,000 5,000 3,996 4,000 4,000 3,688 High exports High imports 3,000 3,000 Low exports Low imports 2,000 2,000 1,000 1,000 0 0 Source: Own elaboration based on Population Census, Integrated Public Use Microdata Series and Customs data. Notes: (1) The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period. (2) A municipality has relatively high exports (imports) if its exports (imports) per 2000 worker ratio in 2015 is above the median across municipalities and low exports (imports) if its ratio lies below the median. (3) Total and average labor incomes are expressed in 2015 Mexican pesos. Fourth, it is also possible that trade induces a shift from informal to more formal work which is not reflected in higher employment rates. We therefore inspect the informality channel and find a strong difference in informality, measured as the percentage of employees without pension or retirement benefits at work for municipalities with high trade per worker relative to those with low trade per worker (Figure 5). In particular, the average informality rate in municipalities with low exports per worker is 13 percentage points higher than in municipalities with high exports per worker (left panel). The difference ascends to almost 19 percentage points for imports (right panel). The scatterplot across all Mexican municipalities suggests a slightly negative correlation between exports and imports per worker and informality (Appendix 4). 6 Figure 4: Exports, imports and migration Exports Imports Number of immigrants from other Mexican immigrants from other Mexican Number of158.3 160 142.1 states 160 states 140 140 120 120 100 100 80 63.3 High exports 80 High imports 60 60 47.1 Low exports Low imports 40 40 20 20 0 0 Number of immigrants from other countries Number of immigrants from other countries 160 160 140 140 120 120 100 100 High exports High imports 80 80 Low exports Low imports 60 42.4 60 43.3 40 24.7 40 23.8 20 20 0 0 Source: Own elaboration based on Population Census, Integrated Public Use Microdata Series and Customs data. Notes: (1) The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period. (2) A municipality has relatively high exports (imports) if its exports (imports) per 2000 worker ratio in 2015 is above the median across municipalities and low exports (imports) if its ratio lies below the median. Figure 5: Exports, imports and informality rate Exports Imports Informality rate Informality rate 100 100 87.9 85.0 80 72.0 80 69.1 60 60 High exports High imports 40 Low exports 40 Low imports 20 20 0 0 Source: Own elaboration based on Population Census, Integrated Public Use Microdata Series and Customs data. Notes: (1) The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period. (2) A municipality has relatively high exports (imports) if its exports (imports) per 2000 worker ratio in 2015 is above the median across municipalities and low exports (imports) if its ratio lies below the median. (3) Labor informality rate is measured as the percentage of employees without pension or retirement benefits at work. In addition, this study also sheds light on the role of geography, sectors and types of trade which can all explain differences in local labor market outcomes. As for trade, we specifically delve into the role of final versus intermediate imports and GVC spillovers. Drawing on Economic Census data, the analysis 7 applies two measures of GVC participation: (i) the share of GVC establishments in a state’s total number of establishments, and (ii) the share of GVC establishments in a state’s total establishments’ employment. Figure 6 shows GVC participation across Mexican states for 2015. The maps clearly indicate a stark geographical concentration of GVC activity in Northern Mexico, close to the U.S. border, which can be explained by geographical distance to key end markets as an important determinant of GVC participation (Fernandes et al. 2022). In Mexico’s case, GVC participation has been driven by the manufacturing sector, with export shares of electrical equipment and electronics, machinery and transport equipment expanding strongly following Mexico’s accession to NAFTA, while services played a smaller role (Winkler et al. 2021). This study therefore disentangles the effects of trade on local labor markets in Mexico’s North and South, and also differentiates between manufacturing and services labor market outcomes. Figure 6: GVC participation across Mexican states Source: Own computations. Data: Economic Census 2015. Note: GVC participants are establishments that simultaneously import and export. GVC = global value chain. 2.2 Specification and instruments We estimate the following equation: , − ,− = + (, − ,− ) +Γ,0 + ,,− (1) where , is the labor market outcome measure in municipality m and year t. ,0 is a vector of variables to control for different trends across municipalities. These include literacy rate, sectoral structure of employment, total population, and the share of the population in rural areas in the baseline year 0 .4 The change in the main variable of interest, , , is obtained as follows: , ,− ln ( ) − ln ( ), (2) ,0 ,0 where , represents annual trade in municipality m, while ,0 denotes the number of workers ages 12 or higher in municipality m in the year 2000. We apply natural logarithms to limit the weight of extreme values because towards the left. to account for zero exports in one or both years a 4We chose the specification of differences between the initial and last year of the period as we are interested in the long-run effects of trade and therefore need to include additional control variables, rather than municipality fixed effects, which is in line with the seminal paper of Autor, Dorn and Hansen (2013) and other China shock literature. 8 constant equal to 1 is added to the trade per worker variable. We assess three types of trade variables – exports, imports and GVC participation. Ordinary least squares (OLS) estimates of may be biased if there are omitted variables correlated both with labor market outcomes and trade. More specifically, municipalities with higher labor incomes or employment may be more likely to attract foreign direct investment and thereby become a trade hub. This would lead to OLS estimates that are upward biased. Conversely, policy incentives that encourage trading firms to locate in areas with lower employment or labor incomes would downward-bias OLS estimates. To overcome this challenge, an instrumental variable (IV) estimation is used to isolate changes in Mexican trade that are not driven by local factors. Following Rodríguez-Castelán et al. (2020), we take advantage of sectoral patterns of other developing-countries’ trade with the United States, excluding Mexico. In particular, we apply the following instrumental variable for exports: ∑ ∗ ,, ∑ ∗ ,,− ln ( ) − ln ( ), (3) ,0 ,0 ∗ where ,, is the predicted level of exports by sector, defined as follows: ∗ ,, = ,,0 x (1 + , ), (4) and , is the growth rate of exports at the sectoral level s from other low- and middle-income countries (excluding Mexico) to the United States between the initial year and year t. Thus, changes in the instrument are not driven by factors at the municipality level, but only by trends in the exports of other low- and middle-income countries to the United States. These Bartik-type instruments are widely used in the empirical literature to estimate the local impacts of trade.5 For imports, we mirror this methodology and focus on imports of other low- and middle-income countries (excluding Mexico) from the US as follows: , ,− , − ,− : ln ( ) − ln ( ) (5) ,0 ,0 ∗ where ,, is the predicted level of imports by sector s, defined as follows: ∗ ′ ,, = ,,0 x (1 + , ) (6) ′ and , is the rate of growth of sector imports of other low- and middle-income countries from the United States between year 0 and year . Note that since our GVC participation measures draw on Mexican firm-level data (for more details, see section 2.3), we are unable to create similar instruments for them. 5For example, Autor, Dorn, and Hanson (2013) estimate the impacts of import competition from China on local US labor markets by combining the initial local distribution of employment across sectors and nationwide changes in imports from China by sector. They instrument this variable using changes in Chinese imports by other high-income countries. 9 2.3 Data and variables We combine different data sources on trade, incomes, labor market outcomes, informality and migration at the municipality level. Export and import data by municipality, year, and sector are obtained from the Tax Authority of Mexico (Servicio de Administración Tributaria) and cover each municipality over the period 2004-14. Data are reported using the 2002 Harmonized System classification at the 4-digit level and converted to the International Standard Industrial Classification (ISIC) (Revision 4) classification using correspondence tables. To construct the instruments, we rely on exports from other low- and middle- income countries (excluding Mexico) to the United States, as well as other low- and middle-income countries’ imports (excluding Mexico) from the United States at the 2-digit industry level from the BACI international trade database (Gaulier and Zignago 2010).6 These countries encompass all low- and middle- income countries, following the World Bank’s income classification of the year under consideration. Figure 7 shows that the association between the instrument and the endogenous variable is strong. Figure 7: Correlation between the instrument and the endogenous variable Exports Imports Source: Own elaboration based on Customs data, IPUMS and BACI. Notes: The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period (2005 and 2015). There are many municipalities with zero exports and imports. Rather than dropping these municipalities, we chose to include them, following Rodríguez-Castelán et al. (2020). Our analysis also draws on the Economic Census data of 2004 and 2014 for two extensions. First, the analysis splits imports into intermediate and final imports, as their impact could vary. Moreover, since exports and imports, in particular intermediate imports, even at the municipality level, are highly correlated, differentiating between intermediate and final imports allows to obtain a cleaner picture of the effect of imports on the local labor market.7 Imported intermediates are defined as the sum of all imported raw materials for production as well as goods or materials used to offer a service. Final imports are goods and services purchased for resale without further transformation. Second, we also compute GVC participation measures, based on establishments that simultaneously import and export (GVC- participating establishments). This has been a widely accepted measure of GVC participation at the firm or plant level (see, for instance, World Bank 2020). Because many municipalities have zero changes in GVC 6 These include the following sectors: agriculture, mining, food products, textiles, wood, paper, printing, chemicals, rubber and plastics, metal, electronics, machinery, automotive, other manufacturing, utilities, construction, education, and other nonmanufacturing. 7 The correlation coefficient between exports per worker and imports per worker (from the Economic Census) is 0.616. The correlation coefficient between exports per worker and intermediate imports per worker is 0.621, while that with final imports per worker is 0.354. 10 participating establishments over the period 2004-14,8 our analysis focuses on GVC participation measures at the state level. We compute two measures that are used in the regressions: the share of GVC establishments in a state’s (i) total number of establishments, and (ii) total establishments’ employment. Correlates with labor market outcomes in that case are to be interpreted as GVC spillovers. To obtain labor market, informality and migration indicators as well as information on the demographic characteristics of the population at the municipality level, we rely on tabulations from the 2000 and 2010 Census of Population and Housing and the 2015 Population Count.9 Given that publicly available data from the Mexican Institute of Statistics do not allow for more than two cross-tabulations, we use census microdata samples from the Integrated Public Use Microdata Series to estimate labor market indicators for more detailed groups (Ruggles et al. 2007). Because the 2005 data do not include detailed information on labor market outcomes, 2000 data are used instead as the initial year for these indicators (with the exception of migration variables). The analysis also uses information on remittances received in Mexico from the United States (CONAPO 2014). The analysis focuses on four types of labor market outcomes, namely (i) employment-related measures (labor force participation rate and employment rate), (ii) labor incomes (total and average), (iii) measures of migration, and (iv) labor informality. First, the employment channel captures both labor force participation rate and employment rates which are based on the IPUMS variable EMPSTAT taking the values of 0 (not in universe10), 1 (Employed), 2 (Unemployed) and 3 (Inactive). The labor force participation rate is defined as the share of employed and unemployed in the universe (sum of employed, unemployed and inactive), whereas the employment rate is the share of employed in the universe. In other words, the difference between labor force participation and employment rates is the share of unemployed in the universe. Second, the income channel uses two measures, namely total and average labor incomes. Total labor income captures a municipality’s total sum of labor incomes, while averages are obtained by dividing total labor income by the total number of employed workers. Third, the migration channel is captured by the number of residents who lived in another state of Mexico or abroad 5 years before (immigration), the number of residents who lived in another state of Mexico 5 years before (domestic immigration) and the number of residents who lived abroad 5 years before (international immigration). The analysis also includes the total working age population (aged 14-55) as outcome variable. Fourth, labor informality rate is measured as the percentage of wage and salary workers without pension or retirement benefits (SAR or AFORE) at work. Finally, the analysis uses three policy measures at the regional level, namely (i) the initial number of kilometers of road network in a municipality, (ii) the initial labor market flexibility of the municipality’s state drawing from Mendoza-Cota (2017) who estimate an index of numerical labor flexibility of the formal labor market,11 and (iii) education spending per student of the municipality’s state.12 We also include several demographic and employment control variables in the regressions to account for different 8 1,619 municipalities out of the 1,980 in the sample. 9 Even though the 2015 population count is from a survey, the sample size is sufficiently large to provide statistics representative at the municipality level. See Enamorado et al. (2016). 10 The universe is the population aged 12 and older. 11 In the absence of state-level measures of legal restrictions and costs of hiring and firing workers for Mexico, we apply a measure of labor flexibility of the formal labor market for the year 2005 (Mendoza-Cota 2017). It is based on information on formal contract workers registered with the Mexican Institute of Social Security (IMSS) which provides health care and pensions. The measure captures the share of temporary urban and rural contract workers in total (i.e., temporary and permanent) contract workers. 12 Education expenditure per student in the state is obtained as the sum of federal, state and municipal education expenditure in the state in 2005 divided by the total number of students enrolled at all levels of education in the state in 2005/2006. 11 trends in trade and labor market outcomes. These include the share of rural population, the literacy rate (share of population aged 15 and older who is literate), total population, and the shares of workers in the primary, secondary and tertiary sectors of economic activity as well as the median income per capita (all variables for the year 2000). The summary statistics are shown in Appendix 5. 3. Trade and local labor market effects: Baseline results 3.1 Exports, imports and the employment channel Table 1 shows the impact of exports and imports on labor force participation and employment rates, using both OLS and the IV approach. The OLS estimates indicate that higher exports per worker tend to increase labor force participation rates (panel a). In particular, a 10 percent increase in exports per worker raised the share of working-age individuals working or looking for a job by 0.05 percentage points (column 6). The economic significance of this impact is rather small, considering that the labor force participation rate in the average municipality in 2000 was 45.8 percentage points. In line with these findings, the OLS results suggest that exports per worker and employment rates are strongly associated, although the economic significance is small (panel b). Specifically, a 10 percent increase in exports per worker is associated with an increase of 0.06 percentage points in the local employment rate, which represents only 0.13 percent of the employment rate of the average municipality in 2005 (see also Rodríguez-Castelán et al. 2020). Because OLS estimates may be biased if unobserved shocks at the local level are correlated with both exports and income levels, the analysis adopts an IV approach. The results of the first-stage regressions indicate that the instrument is highly correlated with the endogenous variable, as the F-statistic exceeds the standard significance threshold (see Appendix 6 and Figure 7). Applying the IV approach, higher exports per worker raise labor force participation rates significantly and the magnitude increases by a factor of three relative to OLS estimates. In contrast, the IV estimates are less significant than OLS estimates for employment rate. The only statistically significant result is the specification excluding control variables, suggesting that a 10 percent increase in exports raised employment rates by 0.24 percentage points on average. Similarly, focusing on imports per worker (Table 1, bottom panels), the IV estimates only find a significant effect on both labor force participation and employment rates when no control variables are added. On the positive side, the absence of negative effects points to a more nuanced role of imports, as the latter could include not only final imports that may compete with domestic production, but also imported inputs used for domestic or export production. Overall, the results so far find more significant and stronger effects for exports than for imports per worker. 12 Table 1: Exports, imports and local jobs OLS IV (a) Impact of exports on labor force participation (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.423*** 0.553*** 0.550*** 2.821** 1.639* 1.664* (0.148) (0.134) (0.134) (1.129) (0.958) (1.000) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact of exports on employment (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.457*** 0.587*** 0.589*** 2.561** 1.354 1.405 (0.151) (0.139) (0.140) (1.126) (0.968) (1.010) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact of imports on labor force participation (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.127 0.0772 0.0783 1.238*** -0.217 -0.220 (0.0846) (0.0762) (0.0762) (0.369) (0.231) (0.230) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact of imports on employment (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.130 0.0816 0.0818 1.296*** -0.144 -0.145 (0.0866) (0.0794) (0.0794) (0.378) (0.244) (0.243) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Note: Robust standard errors in parentheses. *** p < .01 ** p < .05 * p < .1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate and employment rate. The explanatory variable is the change in the log of exports-to- and imports-to-workers workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). 3.2 Exports, imports and the labor income channel Table 2 shows the impacts of exports and imports per worker on labor incomes. The weak impacts of exports on the employment rate in Table 1 may be due to the fact that higher exports per worker led to a net inflow of migrants, thereby increasing both the numerator and the denominator of the employment rate (see section 2.2). We therefore assess the impact on total labor incomes, i.e., the sum of all labor incomes in a municipality. The effect of exports on the total labor income at the local level is statistically significant. A 10 percent increase in exports per worker raised total labor incomes by 2.4 percent on average (panel a, column 6). Considering that total labor incomes in real terms grew by 9.2 percent during the period, this finding is economically significant (see also Rodríguez-Castelán et al. 2020). Focusing on imports per worker, results confirm the positive effect on total labor incomes, although the impact is economically weaker. This could be due to offsetting effects of imports, as the latter could be import-competing (e.g., in the case of final consumer goods) or complementing domestic production. While the former reduces output and labor demand, the latter is linked to enabling and increasing output and employment. However, the average labor income per worker did not grow in response to export and import growth (panels b and d). As was the case for labor force participation and employment rates, imports showed a weaker effect on local labor incomes than exports, both statistically and economically. 13 Table 2: Exports, imports and local labor incomes OLS IV (a) Impact of exports on total labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0216 0.0216 0.0263** -0.00253 0.189** 0.241*** (0.0134) (0.0132) (0.0131) (0.0794) (0.0858) (0.0920) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact of exports on average labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.00405 -0.00488 -0.00205 0.0201 0.0514 0.0802 (0.00742) (0.00746) (0.00745) (0.0497) (0.0532) (0.0575) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact of imports on total labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.00388 0.00754 0.00662 -0.0529* 0.0393* 0.0418* (0.00760) (0.00735) (0.00720) (0.0272) (0.0233) (0.0233) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact of imports on average labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.00249 0.00315 0.00259 -0.0145 0.00482 0.00641 (0.00416) (0.00416) (0.00416) (0.0127) (0.0120) (0.0120) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Note: Robust standard errors in parentheses. *** p < .01 ** p < .05 * p < .1. The dependent variable is the change from 2000 to 2015 in the log of total labor income and log of average labor income. The explanatory variable is the change in the log of exports- to-workers and imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. 3.3 Exports, imports and the migration channel Our findings show that both export and import expansion per worker increase immigration to Mexican municipalities, but effects are stronger for exports (Tables 3 and 4). Expanding exports per worker by 10 percent increased the number of people migrating into the municipality in the previous five years by on average 7.2 people (see also Rodríguez-Castelán et al. 2020), while the effect of imports is less than half that magnitude (panel a, column 1). This represents 4.7 and 1.4 percent of the number of recent immigrants for the average municipality in 2005 for exports and imports, respectively. Domestic immigrants accounted for the majority of trade-induced immigration (panel b) compared with international immigrants. While a 10 percent expansion of exports (imports, respectively) per worker led to a rise in the number of immigrants from other Mexican municipalities by about 5 (2.7, respectively) people, the increase was only 2.2 (0.54, respectively) people for immigrants from other countries (panel c). Exports and imports attracted predominantly unskilled immigrants. The number of unskilled migrants increased by 5.1 people and 2 people, respectively, due to higher exports and imports (panel a, columns 14 2-4). In contrast, the increase in the number of semi-skilled and high-skilled immigrants was substantially lower. Table 3: Exports and local migration patterns IV (a) Impact on the number of immigrants All Unskilled Semi-skilled Skilled log (Exports per worker), change 75.63*** 53.37*** 11.14*** 7.257*** (19.36) (13.32) (3.562) (2.444) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (b) Impact on the number of immigrants from other Mexican states All Unskilled Semi-skilled Skilled log (Exports per worker), change 52.33*** 37.79*** 8.065** 5.874*** (15.88) (11.50) (3.230) (2.196) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (c) Impact on the number of immigrants from other countries All Unskilled Semi-skilled Skilled log (Exports per worker), change 23.30*** 15.58*** 3.072*** 1.383*** (5.832) (3.506) (0.648) (0.475) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (d) Impact on the size of the working age population All Unskilled Semi-skilled Skilled log (Exports per worker), change 1,310*** 873.3*** 306.5*** 123.8*** (273.5) (206.0) (57.87) (27.79) Observations 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2005 to 2015 for the group specified in the column header, namely: the number of residents who lived in another state of Mexico or abroad 5 years before (panel a); the number of residents who lived in another state of Mexico 5 years before (panel b); the number of residents who lived abroad 5 years before (panel c); the total working age population (aged 14-55) (panel d). The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Unskilled: Primary completed or less; Semi-skilled: Secondary completed; Skilled: University completed. Table 5 shifts the focus to migration patterns with the United States, specifically the impact of exports and imports on emigrants to the United States and returning migrants from the United States to Mexico over the period 2000-10.13 Not surprisingly, expanding exports per worker by 10 percent lowers the number of households with emigrants to the United States by about 9 while increasing the number of households with returning migrants from the United States by 14. In other words, larger exports deter outward migration and encourage emigrants to return. As a result of these changing migration patterns, remittances from the United States dropped following an expansion in exports per worker. Specifically, a 10 percent increase in the export-to-worker ratio reduced the share of households receiving remittances from the United States by 0.19 percentage 13 These data are not available for 2005 and 2015. 15 points (column 3). This represents 1.44 times the decline in the share of households receiving remittances over the period. While exports affect remittances through a lower number of emigrants in a municipality, another channel may be that higher labor force participation could reduce reliance on remittances. Interestingly, while growing imports also attract back emigrants from the United States, they did not deter outward migration and had no impact on remittances from the United States. Table 4: Imports and local migration patterns IV (a) Impact on the number of immigrants All Unskilled Semi-skilled Skilled log (Imports per worker), change 34.50*** 21.49*** 6.672*** 3.863*** (7.094) (4.871) (1.130) (0.626) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (b) Impact on the number of immigrants from other Mexican states All Unskilled Semi-skilled Skilled log (Imports per worker), change 28.86*** 19.55*** 5.917*** 3.396*** (5.835) (4.452) (1.032) (0.562) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (c) Impact on the number of immigrants from other countries All Unskilled Semi-skilled Skilled log (Imports per worker), change 5.641*** 1.947** 0.754*** 0.467*** (1.912) (0.958) (0.186) (0.102) Observations 1,980 1,980 1,972 1,868 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES (d) Impact on the size of the working age population All Unskilled Semi-skilled Skilled log (Imports per worker), change 331.8*** 204.9*** 85.28*** 40.32*** (51.42) (35.48) (12.80) (6.836) Observations 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) YES YES YES YES Median income control (year 2000) YES YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2005 to 2015 for the group specified in the column header, namely: the number of residents who lived in another state of Mexico or abroad 5 years before (panel a); the number of residents who lived in another state of Mexico 5 years before (panel b); the number of residents who lived abroad 5 years before (panel c); the total working age population (aged 14-55) (panel d). The explanatory variable is the change in the log of imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Unskilled: Primary completed or less; Semi-skilled: Secondary completed; Skilled: University completed. Domestic and international migration patterns also help explain the positive effect of growing exports per worker on the size of the working-age population aged 14-55 in urban municipalities, especially among the unskilled (Table 3, panel d). A 10 percent expansion in exports per worker raised the skilled working-age population by about 12 people, compared to a staggering 83 people for the unskilled working-age population, accounting for roughly 30 percent of the increase in the number of unskilled people in this age-group in the average municipality over the period 2004-14. The positive effect of higher exports per worker on inward migration suggests that the positive labor market effects are to some extent diluted because of the increase in local labor supply, especially the unskilled. Growing imports per worker 16 also significantly increase the working-age population, although to a lesser extent than exports. The unskilled working-age-population again accounts for the bulk of this increase. Table 5: Exports, imports and local migration patterns with the US IV Emigration Returning migrants Remittances (a) Exports to USA from USA from USA log (Exports per worker), change -95.97** 152.3*** -1.984*** (38.36) (33.16) (0.639) Observations 1,979 1,979 1,979 Demographic and employment controls (year 2000) YES YES YES Median income control (year 2000) YES YES YES Emigration Returning migrants Remittances (b) Imports to USA from USA from USA log (Imports per worker), change -14.59 22.58*** -0.257 (10.19) (7.726) (0.195) Observations 1,979 1,979 1,979 Demographic and employment controls (year 2000) YES YES YES Median income control (year 2000) YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variables are: Emigration to the US = change in the number of households with emigrants to the US from the previous five-year period, from 2000 to 2010; Returning migrants from USA = change in the number of households with returning migrants to the US in the previous five-year period, from 2000 to 2010; Remittances from USA = change in the percentage of households receiving remittances from US, from 2000 to 2010. The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014 in panel (a) and the change in the log of imports-to-workers from 2004 to 2014 in panel (b). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). 3.4 Exports, imports and the informality channel Besides migration, informality could serve as an alternative channel through which trade affects local labor markets in low- and middle-income countries. Even when trade does not change the rate of employed workers, it could induce a movement of workers from the informal to the formal labor market. We therefore estimate the impact of exports and imports on local labor informality, defined as the percentage of employees without pension or retirement benefits at work. Focusing on our preferred specification including the relevant control variables in column (6) highlights that exports significantly reduce local labor informality (panel A). A 10 percent expansion in exports per worker lowered the informality rate by 0.24 percentage points on average. We also observe a decline of labor informality as a result of increasing imports per worker, although the coefficient is much smaller (panel B). A 10 percent expansion in imports per worker lowered informality by 0.07 percentage points on average. 17 Table 6: Exports, imports and local labor informality OLS IV (a) Impact of exports on informality rate (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.338 -0.319 -0.391* -1.444 -1.756 -2.483* (0.236) (0.234) (0.232) (1.327) (1.403) (1.447) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact of imports on informality rate (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.426*** -0.433*** -0.420*** -0.453 -0.739* -0.777* (0.142) (0.138) (0.137) (0.413) (0.413) (0.412) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Note: Robust standard errors in parentheses. *** p < .01 ** p < .05 * p < .1. The dependent variable is the change from 2000 to 2015 in the informality rate, measured as the percentage of employees without pension or retirement benefits at work. The explanatory variable is the change in the log of exports-to-workers and imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). 4. Trade and local labor market effects: Extensions 4.1 Geography: North versus South Geographical distance is a major determinant of GVC participation and Mexico’s states and municipalities in the North are more closely integrated in GVCs (World Bank 2020, and section 4.4). Geographical proximity to the United States could thus play a major role in determining the local labor market effects of trade. Rather than relying on distance measured in kilometers which does not account for trade barriers, we rely on a measure of travel time to one of 44 entry ports to the United States following Blankespoor at al. (2017). We then split municipalities into the North and South, based on a state’s geographical distance to the U.S. border (see Appendix 7), where states in the North (South, respectively) are classified as those below (above, respectively) the median geographical distance across all states. Appendices 8 and 9 show the results for the North and South, respectively. While OLS estimates suggest a positive correlation between rising exports per worker and employment rates and to a lesser extent labor force participation rates in the North, IV estimates find no significant effect. Higher exports per worker in the South, in contrast, significantly increase labor force participation and employment rates. A 10 percent expansion in exports per worker led to an increase in labor force participation and employment rates by 0.25 and 0.22 percentage points, respectively, in the South. The lack of effects through the employment channel in the North is partially explained by the increase in labor supply due to larger net migration compared to the South. In addition, results suggest that higher exports reduce the informality rate only in the South, but not in the North. Specifically, a 10 percent expansion in exports per worker in the South lowered the informality rate by 0.35 percentage points on average. Higher exports per worker, however, raise total labor incomes in the North, but not in the South. Specifically, a 10 percent rise in exports per worker led to gains in total labor incomes by 4.9 percent. There is no effect of higher exports per worker on average labor incomes, though, which again could be explained by increased labor supply especially in the North. The lack of statistical significance on labor incomes in the South may be due to generally fewer job opportunities and lower labor demand. The 18 reduction in the informality rate could also indicate that workers may prefer the benefits of being formal (such as contributions to pensions or retirements) over the prospect of higher incomes. In sum, changes in exports in the South seem to materialize in labor market gains through the employment and informality rather than the income channel. Shifting the focus on imports, the analysis confirms the positive effect of rising imports per worker on total labor incomes in the North, whereas their impact on labor force participation and employment rates are only significant based on the IV estimates when no control variables are added. As was the case for exports, we find a positive effect of imports on net inward migration, partially explaining the lack of effects through the employment channel due to the increase in labor supply, while increasing imports do not affect the informality rate. In the South, rising imports per worker decrease both labor force participation and employment rates. Specifically, a 10 percent growth in imports per worker reduces labor force participation and employment rates by 0.066 and 0.071 percentage points, respectively. It appears that given the lower integration into trade and GVCs in the South, imports are more likely to be competing with domestic production, thereby also lowering the demand for local labor. This finding could also explain why an expansion of imports in the South does not reduce the informality rate. Due to the lower job prospects, our results also find that the effect of exports and imports on immigration is smaller in the South. 4.2 Sectors: Trade and sectoral labor market effects A municipality’s trade expansion could affect labor market outcomes across sectors differently. Appendices 10 and 11 show the results differentiating between manufacturing and services labor market outcomes using OLS and IV estimation. Focusing on the IV estimates, higher exports per worker resulted in higher total manufacturing labor incomes, but not average manufacturing labor incomes (Appendix 10, Panel A). A 10 percent increase in exports per worker thus expanded total labor incomes in the manufacturing sector in Mexico by 2.3 percent. However, the results find no labor market effects through the employment channel, which again could be partially explained by the diluting effect of the migration channel increasing labor supply. Expanding exports per worker also reduces the informality rate in the manufacturing sector. In particular, a 10 percent increase in exports lowers the informality rate by 0.55 percentage points on average. In contrast, increased imports per worker had no significant labor market outcomes in manufacturing via any channel. Interestingly, there is a positive effect of trade expansion at the municipality level for labor market outcomes in services through the income channel (Appendix 11). Higher exports and imports per worker both increase total labor incomes in services, although the effect is more pronounced for exports than for imports. A 10 percent rise in exports per worker expanded total labor incomes in the service sector by 3.2 percent, while a 10 percent increase in imports per worker affected total labor incomes in services by almost 0.5 percent. Overall, our findings suggest that export expansion in Mexico benefitted total labor incomes of both manufacturing and service sectors. Since trade data in our dataset consist mainly of goods, this finding points to an indirect linkage effect, i.e., goods exports may rely on domestic services such as business services or utilities, thereby increasing total labor income also in service sectors. 4.3 Type of imports: Final versus intermediates This section assesses whether the type of imports matters differently for the labor market effects from trade expansion. The absence of significant effects of imports through the employment and income channels in section 3 could point to a more nuanced role of imports. The latter include not only final imports that can compete with domestic production, but also imported inputs used for domestic or export production. 19 Appendix 12 reports the OLS results distinguishing between intermediate and final imports.14 Import data in these regressions draw on the Economic Census data. There are no significant correlations between imports and labor market outcomes when all controls are included. However, the coefficients between final imports and both the labor force participation and employment rates are negative when no controls are included (column 7), while they are insignificant for intermediate imports (column 4). This could point to a more favorable role of intermediate imports relative to final imports for labor market outcomes through the employment channel, suggesting that imported inputs complement domestic or export production, thereby increasing employment. Interestingly, the results find a negative association between the expansion of final imports and inward migration, but no link with intermediate imports. There is no relationship with the informality rate. Differentiating by broad sector of traded imports – manufacturing versus services – reveals that the observed overall pattern is driven by manufacturing imports (Appendix 13). The results are in sharp contrast for services imports (Appendix 14). Here, we find a positive and significant association of both intermediate and final services imports with labor force participation and employment rates. The coefficient between intermediate services imports and employment rates is slightly larger, which could imply that access to intermediate services for domestic or export production has larger employment benefits. Focusing on the income channel, however, the findings are less clear. Both final and intermediate services imports are positively associated with total labor incomes in Mexico, but only the estimates for final services imports are statistically significant. Finally, we find no association between services imports and labor markets through the migration and informality channels. 4.4 Imports for export production: GVC spillovers Finally, we analyze the role of GVC participation for local labor market outcomes. Because many municipalities have zero changes in GVC participating establishments over the period 2004-14, our analysis focuses on GVC participation measures at the more aggregated state level. Correlates with labor market outcomes in that case are to be interpreted as GVC spillovers. Due to the lack of instruments that capture GVC participation at the state level, we can only rely on OLS estimates, as reported in Appendix 15. Our GVC spillover variables measure the share of GVC establishments in a municipality’s (i) total number of establishments, and (ii) total establishments’ employment. Focusing on the employment-based GVC measures (columns 4 to 6), the findings suggest that GVC participation at the state level is positively correlated with labor market outcomes at the municipality level through the employment channel (both for labor force participation and employment rates). We also find a positive association with total labor incomes at the municipality level, but a negative one with average labor incomes. There are positive GVC spillovers for both employment and total labor income at the municipality level. The negative relationship between GVC spillovers and average wages implies a stronger role of the employment channel than the income channel (possibly driven again by increasing labor supply due to inward migration), given that total labor income divided by the number of workers yields average labor incomes. The results do not find any relationship between GVC spillovers and inward migration nor the informality rate. Using the GVC measures based on number of establishments instead confirms most of these patterns, although the results are less significant. This may be due to the fact that establishment-based measures do not account for differences in the economic importance of establishments. 14The F-statistic does not exceed the standard significance threshold, so we cannot use the existing instrument (section 2.1) for the endogenous variable. 20 4.5 Mediating factors: The role of policy Finally, this study explores the role of selected policy factors that mediate the impact of trade on local labor markets, differentiating between connectivity, labor market flexibility and spending on education. In a first step, we focus on connectivity, which has been identified as determinant of trade and GVC participation across a large set of countries (Fernandes, Kee and Winkler 2021). Evidence for Mexico finds that faster access to local markets had a positive impact on local employment (Blankespoor et al. 2017). Our selected measure is the initial road network (in kilometers) in a municipality. Surprisingly, the results suggest that export expansion raises total labor incomes and the number of immigrants more strongly in municipalities with lower initial road networks (Appendix 16). In contrast, higher imports per worker increase total labor incomes and the number of immigrants more strongly in municipalities with higher initial road networks. The latter finding could be linked to imports for export production (GVC participation), as lead firms require good connectivity to be able to deliver their output in a timely and reliable manner. Labor market flexibility determines how quickly firms and workers react to trade shocks in terms of adjustments in the quantity of workers and is thus expected to lead to different labor market outcomes. In line with expectations, we find that the quantity of labor reacts more strongly to export expansion in municipalities with high labor market flexibility,15 as suggested by the larger positive effect on the number of immigrants (Appendix 18). The regressions for imports confirm this result and additionally find larger declines in informality in municipalities with high labor market flexibility. Contrasting these findings, the price of labor, as measured by average labor incomes, reacts more strongly in municipalities with low labor market flexibility, which holds for both exports and imports. Finally, spending on education can shape the quality of labor supply in a region, which, in turn, is a driver of trade and GVC participation (Fernandes, Kee and Winkler 2021). The analysis divides municipalities according to their states’ spending on education (at the federal, state and municipality level) per enrolled student into high and low. In regions with higher spending on education, export expansion benefited labor markets through growing total labor incomes, implying a higher quality of labor. In regions with lower spending on education, the labor market gains from exports materialized through the employment and informality channels, as evidenced by growing labor force participation and employment rates and the declining informality rate (Appendix 19). Interestingly, the migration channel is stronger in regions with lower spending on education, both for exports and imports. It may be possible that the availability of higher-skilled labor in a region acts as a deterrent for inward migration, as trade predominantly attracts unskilled migrants (see section 3.3) and they would find it harder to compete for jobs. Focusing on imports, regions with high spending on education continue to gain through the income channel, but also experience reduced labor force participation and employment rates. It may be possible that the types of imports in regions with larger skills supply compete more strongly with domestic production, thus hurting the employment channel. Imports only affect labor markets through the migration channel in regions with low spending on education. 5. Conclusions and policy implications This paper contributes to the literature on the labor market effects of trade in several ways. First, this analysis provided a framework and new insights into the underlying channels through which trade can affect labor markets and broader development, distinguishing between the employment, labor income, migration, and informality channels. Second, our study contributes to the number of studies 15 See Appendix 17 for the list of states and their classification into high and low labor market flexibility. 21 assessing the impact of trade on labor markets for Mexico specifically. Third, it adds to the scarce number of studies on low- and middle-income countries that investigate the effect of trade on local labor market outcomes. Fourth, the analysis also sheds light on differences by geography and sectors. Finally, our study not only examined the impact of exports but also of imports, especially of different types of imports and GVC spillovers more broadly. First, the study finds that expanding exports per worker in Mexico’s municipalities increased labor force participation, but not employment rates. Exports also raised total labor incomes, but not average labor incomes, implying a growing labor supply due to increased (mainly unskilled) immigration into Mexican municipalities. It also finds that trade induces a shift from informal to more formal work which explains – in addition to migration – the lack of effects through the employment channel. Second, the analysis sheds light on differences by geography and sectors. Trade affected labor markets in the North through the income and migration channel, while it affected municipalities in the South through the employment and informality channel. Export expansion in Mexico benefitted total labor incomes of both manufacturing and service sectors, possibly through an indirect linkage effect of goods exports using domestic services inputs. It also reduced informality in the manufacturing sector. Import expansion had no labor market effects in the manufacturing sector but raised total labor incomes in the service sector. Third, the absence of effects of imports through the employment and income channels points to a more nuanced role of imports which can compete with or complement domestic production. Our findings suggest a more favorable role of intermediate relative to final imports through the employment channel which is driven by manufacturing imports. In contrast, we find a positive and significant role of both intermediate and final services imports through the employment channel. The analysis also finds some evidence for positive spillovers from GVC participation at the state level on labor markets at the municipality level through the employment channel as well as a positive association with total labor incomes. Finally, the analysis examines how policy measures at the state level mediate the labor market effects from trade, focusing on connectivity, labor market flexibility and spending on education. The study finds that higher imports per worker increase total labor incomes and the number of immigrants more strongly in municipalities with higher initial road networks, which could be linked to imports for export production (GVC participation). As for the role of labor market flexibility, results imply that the quantity of labor (measured by migration and the informality rate) reacts more strongly to export shocks in municipalities with high labor market flexibility, while the price of labor (measured by average labor incomes) reacts more strongly in municipalities with low labor market flexibility. The extent of spending on education also shapes the labor market effects from exports, with regions spending more gaining from higher total labor incomes, and regions spending less benefitting from more employment and reduced informality. The migration channel is stronger in regions with lower spending on education, both for exports and imports, implying that the availability of skills in a region acts as a deterrent for immigrants, who are predominantly unskilled. The findings of this paper point to several policy implications. First, the effects of trade on local labor markets are nuanced and depend on the labor market channels, type of trade, geography, and sectors under consideration. Any attempt to address market failures by policy therefore needs to consider the specific context. Second, our findings imply that in a middle-income country like Mexico, workers adjust through movement – migration to other states or movement from the informal to the formal labor market – whereas evidence for industrialized countries suggests that adjustment happens through job loss, movement to other sectors or unemployment (Autor, Dorn and Hanson 2013, 2019). In addition, higher labor market flexibility within Mexico seems to increase such adjustment through movement. Third, the results for Mexico suggest that a region’s initial spending on education – implying an increase in skilled labor supply – determines the channels through which exports affect local labor 22 markets. More spending raises labor incomes following an export expansion, and also seems to deter unskilled migration. One important finding is that even regions with lower spending on education benefited from export expansion in Mexico, but through other channels, namely increased employment, reduced informality, and stronger inward migration. Fourth, the findings for Mexico highlight several opportunities in services, namely positive labor market outcomes in services industries from total exports and positive income effects from services imports. These gains could be further enhanced by removing restrictions on services trade internationally and domestically. Finally, the domestic economy could more fully benefit from GVCs, for instance, by strengthening linkages to domestic supplying industries as well as the South. It appears that improved connectivity matters in augmenting the labor market benefits from imports. Our analysis also identifies several areas for additional research. First, research could explore the linkages between manufacturing activity and backward linkages to domestic service sectors at the local level more explicitly. Second, additional research is warranted to identify the differential impacts of intermediate versus final imports at the local level within a causal framework, as well as the causal effect of GVC participation for local labor market outcomes. Third, a deeper understanding of the role of skills in coping with the local labor market effects of trade is needed. Finally, more research is needed to assess the role of policy in mediating the effect of trade on local labor markets. 23 References Airola, Jim. 2008. “A Regional Analysis of the Impact of Trade and Foreign Direct Investment on Wages in Mexico, 1984–2000.” Review of Development Economics 12 (2): 276-90. Artuc, E., Lopez-Acevedo, G., Robertson, R. and D. Samaan. 2019. Exports to Jobs : Boosting the Gains from Trade in South Asia, South Asia Development Forum, Washington, DC: World Bank, World Bank and International Labour Organization. Autor, David H., David Dorn, and Gordon H. Hanson. 2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review 103 (6): 2121–68. Balsvik, Ragnhild, Sissel Jensen, and Kjell G. Salvanes. 2015. “Made in China, Sold in Norway: Local Labor Market Effects of an Import Shock.” Journal of Public Economics 127 (July): 137–44. Blankespoor, Brian, Theophile Bougna, Rafael Garduno-Rivera, and Harris Selod. 2017. “Roads and the Geography of Economic Activities in Mexico.” World Bank Policy Research Working Paper No. 8226. World Bank, Washington, DC. Caliendo, L., M. Dvorkin, and F. Parro. 2019. “Trade and Labor Market Dynamics: General Equilibrium Analysis of the China Trade Shock.” Econometrica, 87(3): 741-835. Castilho, Marta, Marta Menéndez and Aude Sztulman. 2012. “Trade liberalization, Inequality, and Poverty in Brazilian States.” World Development 40 (4): 821–35. Chiquiar, Daniel. 2008. “Globalization, Regional Wage Differentials, and the Stolper-Samuelson Theorem: Evidence from Mexico.” Journal of International Economics 74 (1): 70–93. CONAPO (Consejo Nacional de Población, National Population Council). 2014. Índice absoluto de intensidad migratoria: México–Estados Unidos, 2000–2010. Índices sociodemográficos. Mexico City: CONAPO. De Loecker, Jan K., Pinelopi Koujianou Goldberg, Amit K. Khandelwal, and Nina Pavcnik. 2016. “Prices, Markups, and Trade Reform.” Econometrica 84 (2): 445–510 Dix-Carneiro, Rafael, and Brian K. Kovak. 2017. “Trade Liberalization and Regional Dynamics.” American Economic Review 107 (10): 2908–46. Dutt, P., Mitra, D. and P. Ranjan (2009), “International trade and unemployment: Theory and cross- national evidence,” Journal of International Economics, 78(1): 32-44 Elbers, Chris, Jean Olson Lanjouw, and Peter F. Lanjouw. 2003. “Micro-Level Estimation of Poverty and Inequality.” Econometrica 71 (1): 355–64. Enamorado, Ted, Luis F. López-Calva, Carlos Rodríguez-Castelán, and Hernán Winkler. (2016). “Income Inequality and Violent Crime: Evidence from Mexico’s Drug War.” Journal of Development Economics 120 (May): 128–43. Erten, Bilge, and Jessica Leight. 2019. “Exporting Out of Agriculture: The Impact of WTO Accession on Structural Transformation in China.” Review of Economics and Statistics. Published ahead of print, June 26. https://www.mitpressjournals.org/doi/abs/10.1162/rest_a_00852. Erten, Bilge, Jessica Leight, and Fiona Tregenna. 2019. “Trade Liberalization and Local Labor Market Adjustment in South Africa.” Journal of International Economics 118 (May): 448–67. Feenstra, Robert C., and Gordon H. Hanson. 1997. “Foreign Direct Investment and Relative Wages: Evidence from Mexico’s Maquiladoras.” Journal of International Economics 42 (3–4): 371–93. Fernandes, Ana, Hiau-Looi Kee, and Deborah Winkler (2022). “Determinants of Global Value Chain Participation: Cross-Country Evidence,” World Bank Economic Review, Volume 36, Issue 2, pp. 329- 360. Frías, Judith A., David S. Kaplan, and Eric Verhoogen. 2012. “Exports and Within-Plant Wage Distributions: Evidence from Mexico.” American Economic Review 102 (3): 435–40. Gaulier, Guillaume, and Soledad Zignago. 2010. “BACI: International Trade Database at the Product-Level; the 1994–2007 Version.” CEPII Working Paper 2010–23 (October), Centre d’Etudes Prospectives et d’Informations Internationales, Paris. 24 Gonzalez-Barrera, Ana, and Jens Manuel Krogstad. 2019. “What We Know about Illegal Immigration from Mexico.” FactTank (blog), June 28. Hakobyan, Shushanik, and John McLaren. 2016. “Looking for Local Labor Market Effects of NAFTA.” Review of Economics and Statistics 98 (4): 728–41. Hollweg, C., D. Lederman, D. Rojas, E. Bulmer (2014). Sticky Feet- How Labor Market Frictions Shape the Impact of International Trade on Jobs and Wages, Washington, DC: The World Bank. Majlesi, Kaveh, and Gaia Narciso. 2018. “International Import Competition and the Decision to Migrate: Evidence from Mexico.” Journal of Development Economics 132 (May): 75–87. McCaig, Brian. 2011. “Exporting Out of Poverty: Provincial Poverty in Vietnam and U.S. Market Access.” Journal of International Economics 85 (1): 102–13. McCaig, Brian, and Nina Pavcnik. 2018. “Export Markets and Labor Allocation in a Low -Income Country.” American Economic Review 108 (7): 1899–1941. Mendez, O. (2015). The effect of Chinese import competition on Mexican local labor markets. North American Journal of Economics and Finance, 34, 364–380. Mendoza-Cota, Eduardo (2017). “Labor flexibility and regional unemployment in Mexico: a panel cointegration analysis,” Economía, Sociedad y Territorio, xvii(53), 35-62. Nicita, Alessandro. 2009. “The Price Effect of Tariff Liberalization: Measuring the Impact on Household Welfare.” Journal of Development Economics 89 (1): 19–27. Stefan Pahl and Marcel P. Timmer (2020). “Do Global Value Chains Enhance Economic Upgrading? A Long View,” The Journal of Development Studies, 56(9): 1683-1705. Pavcnik, Nina. 2017. “The Impact of Trade on Inequality in Developing Countries.” NBER Working Paper 23878 (September), National Bureau of Economic Research, Cambridge, MA. Revenga, Ana. 1997. “Employment and Wage Effects of Trade Liberalization: The Case of Mexican Manufacturing.” Journal of Labor Economics 15 (3): 20–43. Robertson, Raymond. 2007. “Trade and Wages: Two Puzzles from Mexico.” World Economy 30 (9): 1378– 98. Robertson, Raymond, Timothy J. Halliday, Sindhu Vasireddy (2020), “Labour market adjustment to third- party competition: Evidence from Mexico”, World Economy, 43(7), 1977-2006. Robertson, Raymond, Deeksha Kokas, Diego Cardozo, and Gladys Lopez-Acevedo (2020). “Short and Long- Run Labor Market Effects of Developing Country Exports: Evidence from Bangladesh.” Policy Research Working Paper 9176, World Bank, Washington, DC. Robertson, Raymond, Mexico Alberto Vergara Bahena, Deeksha Kokas, and Gladys Lopez-Acevedo (2021). “International Trade and Labor Markets: Evidence from the Arab Republic of Egypt.” Policy Research Working Paper 9668, World Bank, Washington, DC. Rodríguez-Castelán, C., Vazquez, E. and H. Winkler (2020). “Tracing the Local Impacts of Exports on Poverty and Inequality in Mexico.” WB Policy Research Working Paper No. 9459. Ruggles, Steven, Matthew Sobek, J. Trent Alexander, Catherine A. Fitch, Ronald Goeken, Patricia Kelly Hall, Miriam King, and Chad Ronnander. 2007. “Integrated Public Use Microdata Series: Version 3.0.” Machine-readable database. Minneapolis: Minnesota Population Center. Topalova, Petia. 2010. “Factor Immobility and Regional Impacts of Trade Liberalization: Evidence on Poverty from India.” American Economic Journal: Applied Economics 2 (4): 1–41. Verhoogen, Eric A. 2008. “Trade, Quality Upgrading, and Wage Inequality in the Mexican Manufacturing Sector.” Quarterly Journal of Economics 123 (2): 489–530. World Bank. 2019. Systematic Country Diagnostic: Mexico. Washington, DC: World Bank. World Bank. 2020. World Development Report 2020: Trading for Development in the Age of Global Value Chains. Washington, DC: World Bank. 25 Appendices Appendix 1: Correlation between trade and job outcomes, LFP and employment rate Exports Imports Source: Own elaboration based on Customs data, Population Census and Integrated Public Use Microdata Series. Notes: The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period (2005 and 2015). 26 Appendix 2: Correlation between trade and labor incomes Exports Imports Source: Own elaboration based on Customs data, Population Census and Integrated Public Use Microdata Series. Notes: The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period (2005 and 2015). 27 Appendix 3: Correlation between trade and migration Exports Imports Source: Own elaboration based on Customs data, Population Census and Integrated Public Use Microdata Series. Notes: The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period (2005 and 2015). Appendix 4: Correlation between trade and labor informality Exports Imports Source: Own elaboration based on Customs data, Population Census and Integrated Public Use Microdata Series. Notes: The sample is restricted to the urban and semi-urban municipalities with complete data in the whole period (2005 and 2015). 28 Appendix 5: Summary statistics Municipality average Variable 2005 2015 Change Labor force participation rate 45.8 42.8 -3.0 Employment rate 44.8 40.7 -4.2 Total labor income 6810924 7439688 628764 Average labor income 4258.5 4441.8 183.3 Immigrants from other mexican states or countries: All 153.8 136.3 -17.6 Immigrants from other mexican states or countries: Unskilled 95.8 99.6 3.8 Immigrants from other mexican states or countries: Semi-skilled 23.2 24.1 0.9 Immigrants from other mexican states or countries: Skilled 16.7 12.1 -4.6 Immigrants from other mexican states: All 112.5 102.7 -9.8 Immigrants from other mexican states: Unskilled 77.2 72.4 -4.8 Immigrants from other mexican states: Semi-skilled 20.2 19.3 -0.9 Immigrants from other mexican states: Skilled 14.5 10.7 -3.8 Immigrants from other countries: All 41.4 33.6 -7.8 Immigrants from other countries: Unskilled 18.6 27.2 8.6 Immigrants from other countries: Semi-skilled 3.0 4.8 1.8 Immigrants from other countries: Skilled 2.2 1.5 -0.7 Total working age population (aged 14-55) 2785.8 3105.9 320.1 Unskilled working age population (aged 14-55) 2014.2 2301.5 287.3 Semi-skilled working age population (aged 14-55) 474.2 575.7 101.5 Skilled working age population (aged 14-55) 257.9 219.9 -38.0 Number of households with emigrants to the US 440.2 265.1 -175.1 Number of households with returning migrants from the US 90.1 301.1 211.0 Percentage of households receiving remittances from the US 6.4 6.3 -0.1 Informality rate 67.0 78.5 11.5 Exports/Workers 2000 (+1, in logs) 4.2 4.6 0.4 Exports/Workers 2000 (in levels) 2171.6 3325.9 1154.3 Imports/Workers 2000 (+1, in logs) 2.7 2.8 0.1 Imports/Workers 2000 (in levels) 1845.8 2254.3 408.5 Number of municipalities 1980 1980 Covariates (2000 values) Literacy rate 0.82 Population 44064 Share of population in rural areas 0.85 Share of population > 12 years in primary sector 0.44 Share of population > 12 years in secondary sector 0.15 Share of population > 12 years in tertiary sector 0.41 Per capita median income 1232.5 Policy variables (2005 values) Kilometers of road network 176.2 Labor market flexibility index at state level 12.2 Public spending per student at state level 9752.5 Note: Labor force participation, employment, and labor incomes as well as US migration and remittances in 2005 corresponds to the year 2000, while exports in 2005 refer to the year 2004. Similarly, US migration and remittances in 2015 corresponds to the year 2010, while exports and imports in 2015 refer to the year 2014. Per capita incomes of each decile are expressed at 2014 prices, while average and total labor incomes are at 2015 prices (both deflated using the average National Consumer Price Index- base December 2010). 29 Appendix 6: First-stage regressions IV (a) Exports (1) (2) (3) log (Predicted exports per worker), change 1.135*** 1.092*** 1.057*** (0.153) (0.154) (0.154) Observations 1,980 1,980 1,980 F-statistic 55.01 50.46 47.06 Demographic and employment controls (year 2000) NO YES YES Median income control (year 2000) NO NO YES (b) Imports (1) (2) (3) log (Predicted imports per worker), change 0.768*** 0.828*** 0.830*** (0.0570) (0.0622) (0.0622) Observations 1,980 1,980 1,980 F-statistic 181.5 177.5 177.9 Demographic and employment controls (year 2000) NO YES YES Median income control (year 2000) NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The explanatory variable is the change in the log of the predicted exports-to-workers from 2004 to 2014 in panel (a) and the change in the log of the predicted imports-to-workers from 2004 to 2014 in panel (b) (see equations (3) - (6) in the main text). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). 30 Appendix 7: Average state characteristics Plant-based GVC Employment-based Distance from US State North/South participation GVC participation border (minutes) measure (%) measure (%) Aguascalientes 631 North 0.09% 8.2% Baja California 48 North 0.09% 2.2% Baja California Sur 1127 South 0.02% 0.6% Campeche 1444 South 0.01% 0.0% Coahuila de Zaragoza 233 North 0.12% 8.2% Colima 895 South 0.01% 0.2% Chiapas 1212 South 0.01% 1.0% Chihuahua 201 North 0.08% 2.6% Ciudad de México 866 South 0.09% 2.3% Durango 579 North 0.04% 1.4% Guanajuato 684 North 0.08% 5.2% Guerrero 963 South 0.00% 0.1% Hidalgo 707 North 0.03% 3.6% Jalisco 758 North 0.07% 3.4% México 820 South 0.07% 6.1% Michoacán de Ocampo 848 South 0.02% 1.9% Morelos 871 South 0.05% 5.7% Nayarit 894 South 0.01% 0.1% Nuevo León 162 North 0.26% 7.2% Oaxaca 1074 South 0.00% 0.7% Puebla 907 South 0.04% 6.4% Querétaro 697 North 0.24% 12.3% Quintana Roo 1594 South 0.02% 0.3% San Luis Potosí 562 North 0.10% 9.3% Sinaloa 712 North 0.01% 0.3% Sonora 226 North 0.04% 1.8% Tabasco 1106 South 0.02% 0.7% Tamaulipas 314 North 0.05% 1.8% Tlaxcala 893 South 0.07% 7.6% Veracruz de Ignacio de la Llave 786 North 0.02% 2.0% Yucatán 1633 South 0.03% 1.7% Zacatecas 508 North 0.01% 2.9% Source: Own elaboration based on Blankespoor et al. (2017) and Economic Census 2014. Note: Distance to the U.S. border measures the minimum travel time (in minutes) to one of 44 entry ports to the United States. 31 Appendix 8: Exports, imports and local labor market effects in the North (A) Exports OLS IV (a) Impact on labor force participation (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.296 0.291* 0.284 1.568 0.0841 -0.0471 (0.195) (0.173) (0.174) (1.470) (1.400) (1.478) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on employment (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.404* 0.399** 0.390** 1.552 -0.0154 -0.169 (0.206) (0.184) (0.185) (1.503) (1.438) (1.522) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0334* 0.0392** 0.0409** 0.180 0.438*** 0.491*** (0.0185) (0.0178) (0.0178) (0.133) (0.165) (0.181) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0139** 0.0138** 0.0149** 0.0349 0.103 0.125 (0.00644) (0.00627) (0.00628) (0.0681) (0.0738) (0.0796) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (e) Impact on number of immigrants (1) (2) (3) (4) (5) (6) log (Exports per worker), change 11.13 11.83** 13.13** -222.3** 94.38** 119.8** (7.875) (5.048) (5.216) (88.71) (48.09) (55.64) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (f) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.597* -0.755** -0.859*** 1.996 0.549 -0.999 (0.312) (0.304) (0.309) (2.248) (2.207) (2.181) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES 32 (B) Imports OLS IV (a) Impact on labor force participation (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.203 0.247** 0.250** 1.114*** 0.244 0.236 (0.124) (0.115) (0.114) (0.413) (0.315) (0.312) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on employment (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.244* 0.289** 0.292** 1.292*** 0.431 0.422 (0.127) (0.119) (0.119) (0.430) (0.333) (0.329) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.0116 0.0131 0.0126 0.0278 0.0907*** 0.0917*** (0.0103) (0.00992) (0.00989) (0.0329) (0.0300) (0.0303) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.00433 0.00473 0.00444 -0.0129 0.0127 0.0134 (0.00443) (0.00444) (0.00445) (0.0160) (0.0148) (0.0150) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (e) Impact on number of immigrants (1) (2) (3) (4) (5) (6) log (Imports per worker), change -2.636 4.969** 4.601** -77.40*** 40.55*** 41.46*** (2.718) (1.941) (1.887) (19.29) (10.05) (10.19) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (f) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.649*** -0.544*** -0.514*** -0.799 -0.749 -0.829 (0.195) (0.187) (0.186) (0.551) (0.539) (0.530) Observations 839 839 839 839 839 839 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014 in Table (A) and the change in the log of imports-to-workers from 2004 to 2014 in Table (B). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. See Appendix 7 for the classification of states and therefore municipalities into the North and South. 33 Appendix 9: Exports, imports and local labor market effects in the South (A) Exports OLS IV (a) Impact on labor force participation (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.533** 0.809*** 0.802*** 3.009** 2.660** 2.664** (0.220) (0.213) (0.215) (1.498) (1.301) (1.326) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on employment (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.498** 0.775*** 0.774*** 2.647* 2.295* 2.321* (0.221) (0.217) (0.218) (1.487) (1.304) (1.330) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0119 0.00800 0.0109 -0.0274 0.125 0.142 (0.0178) (0.0177) (0.0178) (0.0976) (0.107) (0.111) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.0208 -0.0196 -0.0167 0.0158 0.0478 0.0642 (0.0127) (0.0128) (0.0128) (0.0664) (0.0744) (0.0772) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (e) Impact on number of immigrants (1) (2) (3) (4) (5) (6) log (Exports per worker), change 15.58* 2.042 2.380 -62.82** 60.50*** 63.49*** (9.022) (7.485) (7.524) (30.19) (17.38) (17.88) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (f) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.101 0.0371 0.0295 -3.450** -3.537* -3.651* (0.344) (0.342) (0.341) (1.716) (1.879) (1.909) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES 34 (B) Imports OLS IV (a) Impact on labor force participation (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.0613 -0.152 -0.153 1.243** -0.693** -0.699** (0.115) (0.0995) (0.0997) (0.622) (0.312) (0.312) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on employment (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.0217 -0.192* -0.192* 1.173* -0.744** -0.747** (0.116) (0.101) (0.102) (0.622) (0.315) (0.315) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.00777 -0.00370 -0.00347 -0.125*** -0.0192 -0.0176 (0.0106) (0.0104) (0.0103) (0.0366) (0.0287) (0.0282) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.000280 0.00121 0.00145 -0.0152 -0.00209 -0.000413 (0.00733) (0.00733) (0.00734) (0.0199) (0.0189) (0.0185) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (e) Impact on number of immigrants (1) (2) (3) (4) (5) (6) log (Imports per worker), change -2.228 1.666 1.693 -99.41*** 24.32*** 24.51*** (3.059) (2.651) (2.668) (26.71) (7.222) (7.296) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (f) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.167 -0.202 -0.202 -0.112 -0.661 -0.666 (0.203) (0.198) (0.198) (0.607) (0.637) (0.638) Observations 1,141 1,141 1,141 1,141 1,141 1,141 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014 in Table (A) and the change in the log of imports-to-workers from 2004 to 2014 in Table (B). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. See Appendix 7 for the classification of states and therefore municipalities into the North and South. 35 Appendix 10: Total exports and local labor market effects in manufacturing (A) Exports OLS IV (a) Impact on employment (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.289*** 0.150** 0.146** 0.605* 0.181 0.149 (0.0781) (0.0691) (0.0696) (0.353) (0.328) (0.337) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0456** 0.0301* 0.0341* 0.0134 0.182 0.226* (0.0180) (0.0179) (0.0180) (0.106) (0.114) (0.119) Observations 1,973 1,973 1,973 1,973 1,973 1,973 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.00536 0.000303 0.00345 0.0305 0.0458 0.0776 (0.00728) (0.00749) (0.00762) (0.0593) (0.0643) (0.0662 Observations 1,973 1,973 1,973 1,973 1,973 ) 1,973 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Exports per worker), change -1.198*** -1.235*** -1.263*** -4.191* -5.252* -5.742** (0.394) (0.394) (0.395) (2.524) (2.753) (2.916) Observations 1,936 1,936 1,936 1,936 1,936 1,936 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES 36 (B) Imports OLS IV (a) Impact on employment (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.109** 0.112*** 0.113*** -0.254** 0.0526 0.0501 (0.0441) (0.0387) (0.0387) (0.114) (0.0934) (0.0929) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.00625 0.0112 0.0105 -0.123*** 0.0264 0.0285 (0.0101) (0.00987) (0.00978) (0.0393) (0.0346) (0.0348) Observations 1,973 1,973 1,973 1,973 1,973 1,973 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.00204 -0.00125 -0.00189 -0.0295* -0.000245 0.00157 (0.00429) (0.00431) (0.00429) (0.0151) (0.0145) (0.0147) Observations 1,973 1,973 1,973 1,973 1,973 1,973 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.441 -0.448* -0.444 -0.0912 -0.534 -0.547 (0.270) (0.271) (0.271) (0.819) (0.815) (0.814) Observations 1,936 1,936 1,936 1,936 1,936 1,936 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the employment rate (panel a), log of total labor income (panel b), log of average labor income (panel c), and informality rate (panel d), all in the manufacturing sector. The labor force participation and migration measures cannot be attributed to specific sectors, because they include working, non-working and inactive people. The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014 in panel (A) and the change in the log of imports-to-workers from 2004 to 2014 in panel (B). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. 37 Appendix 11: Total exports and local labor market effects in services (A) Exports OLS IV (a) Impact on employment (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.0463 0.0237 0.0459 -0.251 0.127 0.337 (0.106) (0.1000) (0.0996) (0.547) (0.549) (0.573) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.0113 0.0140 0.0178 0.0254 0.273*** 0.320*** (0.0144) (0.0139) (0.0139) (0.0856) (0.0959) (0.103) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Exports per worker), change -0.00288 -0.00228 -0.000126 -0.00654 0.0124 0.0330 (0.00756) (0.00752) (0.00751) (0.0492) (0.0525) (0.0553) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Exports per worker), change 0.123 -0.0332 -0.0698 0.125 0.0275 -0.309 (0.215) (0.218) (0.219) (1.429) (1.514) (1.581) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES 38 (B) Imports OLS IV (a) Impact on employment (1) (2) (3) (4) (5) (6) log (Imports per worker), change 0.0728 0.0711 0.0667 0.0110 -0.0878 -0.0753 (0.0581) (0.0575) (0.0567) (0.178) (0.167) (0.165) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (b) Impact on total labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.00237 0.00227 0.00155 -0.0668** 0.0465** 0.0485** (0.00792) (0.00761) (0.00745) (0.0287) (0.0230) (0.0229) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (c) Impact on average labor income (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.00101 -0.000811 -0.00124 0.000802 0.00603 0.00723 (0.00395) (0.00393) (0.00390) (0.0136) (0.0131) (0.0131) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES (d) Impact on informality rate (1) (2) (3) (4) (5) (6) log (Imports per worker), change -0.247* -0.234* -0.227* -0.747 -0.174 -0.194 (0.134) (0.133) (0.133) (0.460) (0.439) (0.438) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls (year 2000) NO YES YES NO YES YES Median income control (year 2000) NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the employment rate (panel a), log of total labor income (panel b), log of average labor income (panel c), and informality rate (panel d), all in the service sector. The labor force participation and migration measures cannot be attributed to specific sectors, because they include working, non-working and inactive people. The explanatory variable is the change in the log of exports-to-workers from 2004 to 2014 in panel (A) and the change in the log of imports-to-workers from 2004 to 2014 in panel (B). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. 39 Appendix 12: Final vs. intermediate imports and local labor market outcomes, OLS Total imports Intermediate imports Final imports (a) Labor force participation (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change -1.221* 0.903* 0.921* -0.875 0.729 0.747 -2.186*** -0.000887 0.00972 (0.714) (0.521) (0.522) (0.989) (0.693) (0.694) (0.640) (0.996) (1.000) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (b) Employment (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change -1.319* 0.803 0.808 -0.933 0.661 0.666 -2.271*** -0.0364 -0.0340 (0.749) (0.543) (0.544) (1.039) (0.723) (0.724) (0.635) (0.965) (0.967) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (c) Total labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change 0.243** 0.0859 0.0707 0.157 0.0500 0.0351 0.272* 0.0958 0.0865 (0.106) (0.0776) (0.0756) (0.105) (0.0921) (0.0913) (0.157) (0.101) (0.0939) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (d) Average labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change 0.0126 -0.0155 -0.0249 -0.0130 -0.0318 -0.0410 0.0944 0.0650 0.0594 (0.0375) (0.0364) (0.0352) (0.0543) (0.0538) (0.0526) (0.0736) (0.0716) (0.0674) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (e) Number of immigrants (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change 240.2*** -1.235 -4.044 201.7** 34.05 31.34 175.1 -74.89** -76.61** (82.79) (49.20) (49.77) (98.12) (69.23) (69.97) (111.1) (34.73) (34.51) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (f) Informality rate (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Imports per worker), change -0.292 0.0796 0.310 0.0808 0.386 0.610 -0.173 0.322 0.461 (1.757) (1.835) (1.836) (2.603) (2.709) (2.694) (0.912) (0.941) (0.898) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the log of imports-to-workers from 2004 to 2014, using total imports in specifications (1) to (3), intermediate imports in specifications (4) to (6) and final imports (obtained as the difference between total and intermediate imports) in specifications (7) to (9). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index, while imports are in 2012's pesos. 40 Appendix 13: Final vs. intermediate manufacturing imports and local labor market outcomes, OLS Total imports Intermediate imports Final imports (a) Labor force participation (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) -1.021 0.485 0.493 -0.720 0.593 0.603 -1.972*** -0.620 -0.618 change (0.660) (0.508) (0.508) (0.904) (0.681) (0.682) (0.619) (0.884) (0.886) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year Median 2000) income control (year 2000) NO NO YES NO NO YES NO NO YES (b) Employment (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) -1.137 0.372 0.371 -0.789 0.515 0.514 -2.074*** -0.666 -0.666 change (0.697) (0.534) (0.534) (0.954) (0.713) (0.715) (0.610) (0.858) (0.857) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year Median 2000) income control (year 2000) NO NO YES NO NO YES NO NO YES (c) Total labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) 0.156 0.0443 0.0320 0.149 0.0428 0.0275 0.121 0.0264 0.0235 change (0.0963) (0.0749) (0.0733) (0.105) (0.0916) (0.0907) (0.119) (0.0872) (0.0807) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year Median 2000) income control (year 2000) NO NO YES NO NO YES NO NO YES (d) Average labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) -0.00482 -0.0205 -0.0270 -0.0193 -0.0335 -0.0417 0.0728 0.0610 0.0595 change (0.0382) (0.0370) (0.0356) (0.0547) (0.0540) (0.0526) (0.0738) (0.0720) (0.0681) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (e) Number of immigrants (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) 140.6* -14.82 -17.35 198.8** 32.12 28.99 -13.26 -104.7*** -105.2*** change (73.17) (51.04) (51.67) (97.99) (69.49) (70.34) (49.87) (33.35) (33.25) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year Median 2000) income control (year 2000) NO NO YES NO NO YES NO NO YES (f) Informality rate (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Manufacturing imports per worker) -0.395 0.0685 0.212 -0.121 0.241 0.419 -0.134 0.418 0.452 change (1.782) (1.852) (1.847) (2.617) (2.716) (2.703) (0.855) (0.863) (0.814) Observations 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 1,830 Demographic and employment controls NO YES YES NO YES YES NO YES YES (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the log of imports-to-workers from 2004 to 2014, using total imports in specifications (1) to (3), intermediate imports in specifications (4) to (6) and final imports (obtained as the difference between total and intermediate imports) in specifications (7) to (9). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index, while imports are in 2012's pesos. 41 Appendix 14: Final vs. intermediate services imports and local labor market outcomes, OLS Total imports Intermediate imports Final imports (a) Labor force participation (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 4.511 23.65** 24.05** 13.42 26.09** 26.57** 3.874 26.37** 26.79** change (5.307) (4.557)* * (4.587) (15.13) (11.69) (11.79) (6.096) (5.094) * * (5.142) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (b) Employment (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 5.167 24.18** 24.38** 16.45 29.19** 29.33** 4.313 26.62** 26.81** change (5.446) (4.652)* * (4.688) (15.36) (12.16) (12.20) (6.239) (5.116) * * (5.166) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (c) Total labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 5.255*** 2.598** 2.369** 4.009 3.051 2.673 6.035*** 2.884** 2.638** change (0.957) (0.464)* * (0.493) (2.545) (2.639) (2.671) (1.113) (0.463) * * (0.486) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (d) Average labor income (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 0.779*** 0.361 0.213 1.394 1.222 0.988 0.784*** 0.277 0.118 change (0.299) (0.265) (0.295) (1.569) (1.595) (1.670) (0.268) (0.219) (0.253) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (e) Number of immigrants (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 5,663*** 936.3* 896.7 909.2 649.4 579.9 6,891*** 1,095 1,052 change (1,681) (560.9) (557.5) (1,273) (811.0) (795.5) (2,084) (682.7) (679.1) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES (f) Informality rate (1) (2) (3) (4) (5) (6) (7) (8) (9) log (Services imports per worker) 4.474 2.056 5.721 65.45*** 70.12*** 75.98*** -2.071 -6.347 -2.469 change (10.84) (11.89) (12.50) (20.83) (17.48) (17.09) (12.89) (14.36) (15.08) Observations 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment NO YES YES NO YES YES NO YES YES controls (year 2000) Median income control (year 2000) NO NO YES NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the log of imports-to-workers from 2004 to 2014, using total imports in specifications (1) to (3), intermediate imports in specifications (4) to (6) and final imports (obtained as the difference between total and intermediate imports) in specifications (7) to (9). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index, while imports are in 2012's pesos. 42 Appendix 15: State-level GVC participation and local labor market outcomes, OLS As share of establishments As share of employment (a) Labor force participation (1) (2) (3) (4) (5) (6) GVC participation, change -51.25*** 8.185 8.761 -0.228** 0.518*** 0.524*** (6.021) (5.285) (5.364) (0.112) (0.0948) (0.0965) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES (b) Employment (1) (2) (3) (4) (5) (6) GVC participation, change -51.08*** 8.750 8.999* -0.218* 0.532*** 0.535*** (6.059) (5.379) (5.458) (0.113) (0.0967) (0.0982) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES (c) Total labor income (1) (2) (3) (4) (5) (6) GVC participation, change 4.090*** 3.182*** 2.857*** 0.0356*** 0.0237** 0.0204** (0.593) (0.588) (0.584) (0.0104) (0.00979) (0.00970) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES (d) Average labor income (1) (2) (3) (4) (5) (6) GVC participation, change 0.851*** 0.701** 0.479 -0.0115** -0.0154*** -0.0176*** (0.277) (0.290) (0.292) (0.00586) (0.00579) (0.00580) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES (e) Number of immigrants (1) (2) (3) (4) (5) (6) GVC participation, change 1,849*** -72.14 -143.2 9.298 -4.850 -5.503 (639.7) (337.8) (344.6) (7.183) (4.612) (4.768) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES (f) Informality rate (1) (2) (3) (4) (5) (6) GVC participation, change -13.10 9.916 15.80* -0.356** -0.0871 -0.0359 (8.528) (8.864) (8.883) (0.154) (0.160) (0.160) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls NO YES YES NO YES YES (year Median2000) income control (year 2000) NO NO YES NO NO YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the labor force participation rate (panel a), employment rate (panel b), log of total labor income (panel c), log of average labor income (panel d), number of immigrants (panel e) and informality rate (panel f). The explanatory variable is the change in the GVC participation from 2004 to 2014, using the share of GVC participants within state in specifications (1) to (3) and the share of GVC participants employment within state in specifications (4) to (6). Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. 43 Appendix 16: Impact of exports and imports on local labor markets, by road density (A) Exports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Exports per worker), change 1.664* 1.405 0.241*** 0.0802 75.63*** -2.483* (1.000) (1.010) (0.0920) (0.0575) (19.36) (1.447) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high road network log (Exports per worker), change -0.269 -0.291 0.298** -0.0336 44.27 -1.330 (1.325) (1.379) (0.116) (0.0720) (28.35) (1.796) Observations 744 744 744 744 744 744 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low road network log (Exports per worker), change 2.098* 1.877 0.327** 0.101 57.75* -1.634 (1.269) (1.293) (0.137) (0.0758) (30.90) (2.092) Observations 742 742 742 742 742 742 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (B) Imports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Imports per worker), change -0.220 -0.145 0.0418* 0.00641 34.50*** -0.777* (0.230) (0.243) (0.0233) (0.0120) (7.094) (0.412) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high road network log (Imports per worker), change 0.0775 0.154 0.0642** -0.0156 40.15*** -0.183 (0.296) (0.322) (0.0312) (0.0120) (8.746) (0.463) Observations 744 744 744 744 744 744 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low road network log (Imports per worker), change -0.371 -0.315 0.0332 0.0320 24.93** -1.150 (0.373) (0.388) (0.0314) (0.0237) (11.01) (0.717) Observations 742 742 742 742 742 742 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the variable specified in the head of the column, except for the migration variable (column 5) for which the change is between 2005 and 2015. The explanatory variable is the change in the log of exports-to-workers and imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. Municipalities with relatively high (low) initial number of kilometers of road network are those municipalities with kilometers of road network in 2005 above (below) the median at municipality level. 44 Appendix 17: Average policy variables across states Spending on Spending on Labor market flexibility Labor market flexibility: State education per education: measure high vs. low student high vs. low Aguascalientes 8.71 Low $ 8,772.25 Low Baja California 7.25 Low $ 13,778.44 High Baja California Sur 20.72 High $ 17,781.01 High Campeche 21.59 High $ 17,367.07 High Coahuila de Zaragoza 8.42 Low $ 12,858.38 High Colima 15.98 High $ 17,782.01 High Chiapas 11.31 Low $ 10,001.61 Low Chihuahua 5.47 Low $ 12,622.13 High Ciudad de México 8.76 Low - - Durango 7.31 Low $ 12,337.00 High Guanajuato 10.30 Low $ 8,417.35 Low Guerrero 19.28 High $ 11,344.63 High Hidalgo 16.17 High $ 6,089.49 Low Jalisco 9.67 Low $ 8,826.69 Low México 13.11 High $ 8,133.54 Low Michoacán de Ocampo 9.77 Low $ 11,788.16 High Morelos 11.69 High $ 8,471.88 Low Nayarit 25.13 High $ 11,451.20 High Nuevo León 8.39 Low $ 9,125.34 Low Oaxaca 12.34 High $ 8,107.83 Low Puebla 10.86 Low $ 7,540.21 Low Querétaro 16.93 High $ 9,854.09 Low Quintana Roo 15.01 High $ 10,476.32 Low San Luis Potosí 14.30 High $ 10,692.10 Low Sinaloa 14.24 High $ 9,781.43 Low Sonora 15.19 High $ 11,685.44 High Tabasco 16.40 High $ 12,827.60 High Tamaulipas 10.31 Low $ 10,764.02 Low Tlaxcala 9.69 Low $ 10,868.01 High Veracruz de Ignacio de la 16.22 High $ 11,029.58 High Llave Yucatán 7.06 Low $ 10,806.38 High Zacatecas 9.49 Low $ 11,742.70 High Source: Own elaboration based on labor market flexibility indexes in 2005 from Mendoza-Cota (2017), total education expenditure at state level in 2005 from the Secretary of Public Education in Mexico (Statistical Reports of the State Educational Financing Questionnaire) and total enrollment in all levels of education in the cycle 2005/2006 from INEGI. Note: States with high (low) labor market flexibility are those with Labor flexibility index in 2005 above (below) the median at the state level and those with high (low) spending on education are those with spending on education per student in 2005 above (below) the median at the state level. Spending on education in ciudad de México is not included due to lack of data of expenditure from national sources. 45 Appendix 18: Impact of exports and imports on local labor markets, by labor market flexibility (A) Exports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Exports per worker), change 1.664* 1.405 0.241*** 0.0802 75.63*** -2.483* (1.000) (1.010) (0.0920) (0.0575) (19.36) (1.447) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year 2000) Median income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high labor market flexibility log (Exports per worker), change 2.306 1.934 0.127 0.0625 80.87*** -3.222 (1.564) (1.568) (0.116) (0.0644) (24.98) (2.137) Observations 1,002 1,002 1,002 1,002 1,002 1,002 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low labor market flexibility log (Exports per worker), change 0.985 0.800 0.321** 0.0903 68.76** -1.680 (1.212) (1.248) (0.145) (0.0954) (30.55) (2.000) Observations 978 978 978 978 978 978 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (B) Imports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Imports per worker), change -0.220 -0.145 0.0418* 0.00641 34.50*** -0.777* (0.230) (0.243) (0.0233) (0.0120) (7.094) (0.412) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high labor market flexibility log (Imports per worker), change -0.170 -0.0595 -0.0224 -0.0215 53.34*** -1.943* (0.469) (0.471) (0.0500) (0.0211) (16.56) (1.027) Observations 1,002 1,002 1,002 1,002 1,002 1,002 Demographic and employment controls YES YES YES YES YES YES (year 2000) Median income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low labor market flexibility log (Imports per worker), change -0.215 -0.157 0.0571** 0.0114 25.19*** -0.447 (0.267) (0.287) (0.0251) (0.0135) (6.444) (0.412) Observations 978 978 978 978 978 978 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the variable specified in the head of the column, except for the migration variable (column 5) for which the change is between 2005 and 2015. The explanatory variable is the change in the log of exports-to-workers and imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. Municipalities with relatively high (low) initial labor market flexibility are those municipalities with Labor flexibility numerical index in the state in 2005 above (below) the median at the state level. Labor market flexibility numerical indexes come from Mendoza-Cota (2017). 46 Appendix 19: Impact of exports and imports on local labor markets, by spending on education (A) Exports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Exports per worker), change 1.664* 1.405 0.241*** 0.0802 75.63*** -2.483* (1.000) (1.010) (0.0920) (0.0575) (19.36) (1.447) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year 2000) Median income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high education spending log (Exports per worker), change 0.0370 -0.271 0.286*** 0.0311 73.64*** 0.322 (1.044) (1.079) (0.103) (0.0548) (28.11) (1.502) Observations 785 785 785 785 785 785 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low education spending log (Exports per worker), change 3.092* 2.901* 0.192 0.135 93.77*** -5.755** (1.715) (1.717) (0.153) (0.106) (30.42) (2.669) Observations 1,188 1,188 1,188 1,188 1,188 1,188 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (B) Imports Labor force Employment Total labor Average Number of Informality (a) Total participation rate income labor income immigrants rate log (Imports per worker), change -0.220 -0.145 0.0418* 0.00641 34.50*** -0.777* (0.230) (0.243) (0.0233) (0.0120) (7.094) (0.412) Observations 1,980 1,980 1,980 1,980 1,980 1,980 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (b) Municipalities with high education spending log (Imports per worker), change -0.488* -0.487* 0.0425* 0.00586 29.56*** -0.700 (0.272) (0.289) (0.0257) (0.0142) (10.37) (0.560) Observations 785 785 785 785 785 785 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES (c) Municipalities with low education spending log (Imports per worker), change 0.0714 0.216 0.0406 0.00644 38.41*** -0.757 (0.345) (0.359) (0.0392) (0.0193) (8.588) (0.608) Observations 1,188 1,188 1,188 1,188 1,188 1,188 Demographic and employment controls YES YES YES YES YES YES (year Median2000) income control (year 2000) YES YES YES YES YES YES Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the change from 2000 to 2015 in the variable specified in the head of the column, except for the migration variable (column 5) for which the change is between 2005 and 2015. The explanatory variable is the change in the log of exports-to-workers and imports-to-workers from 2004 to 2014. Demographic and employment controls include the proportion of rural population, the proportion of population aged 15 and more who are literate, total population, and the share of employment in the primary, secondary and tertiary sectors of economic activity (all variables in the year 2000). Total and average labor incomes are expressed in 2015 Mexican pesos using the National Consumer Price Index. Municipalities with relatively high (low) initial education expenditure are those municipalities with education expenditure per student in the state in 2005 above (below) the median at the state level. Education expenditure per student is obtained as the sum of public education expenditure divided by the total number of students enrolled. 47