Gender, Informal Employment and Trade Liberalization in Mexico

Gender, Informal Employment and Trade Liberalization in Mexico Sarra Ben Yahmed and Pamela Bombarda February 15, 2016 Preliminary Draft Abstract In this paper, we investigate the impact of the NAFTA on male and female employment in formal and informal jobs in Mexico over the period 1993-2001. Using information on Mexican and U.S. import tariffs together with individual information from the Mexican labour force survey, we find that market integration is significantly related to an increase in formal wage employment relative to informal wage employment or self-employment in the tradable sector. Exploiting within-industry variations in tariffs, we show that the formalization of employment occurs within 4-digit manufacturing sectors. By exploiting regional variation in exposure to trade integration, we also explore the effect of trade on reallocation across formal and informal jobs between sectors, including non-tradable sectors, within Mexican municipalities. We find that both men and women working in municipalities exposed to stronger falls in import tariff are more likely to hold informal salaried jobs in the non-tradable sector but that women are more likely to hold formal jobs in the tradable sectors. JEL classification: F13, F16, J16, J21, O15, O17 Keywords: Informality, trade liberalization, gender, Mexico. Acknowledgements: We would like to thank T. Verdier, E. Gamberoni for helpful comments and suggestions. We thank M. Bosch, and M. Manacorda for helping us with the ENEU data and E. Verhoogen for sharing with us the correspondence between Mexican industry classification and ISIC. All errors are ours. This research has been conducted as part of the project Labex MME-DII (ANR11-LBX-0023-01). Corresponding author: Cergy-Pontoise University (THEMA), 33 Boulevard du Port F-95011. Tel.: +33 1 34 25 67 59; fax: +33 1 34 25 62 33. E-mail address: pamela.bombarda@u-cergy.fr 1

1 Introduction This paper documents how trade liberalization has affected gender sorting across types of employment in Mexico during the period 1993-2001. In particular, we provide empirical evidence on how trade integration between Mexico and the US has shaped the repartition of men and women across three types of employment: formal and informal wage employment and self-employment. We exploit both industry and regional variations in tariff exposure to analyse the effect of trade integration on reallocation across those three occupations within and between sectors. We explore how trade liberalization affects gender employment differences in the Mexican labour market between 1993 and 2001. This period is characterized by a dramatic increase in international trade facilitated by the ongoing liberalization efforts that started in the late 1980s. A particular role was played by the NAFTA agreement, which entered into force in January 1994. The agreement provided a gradual tariff liberalization that took place particularly in the first ten years since the entry in force of the treaty. While some authors suggest that anticipation effects were present in the economy, the period 1990-2000 was characterized by an acceleration of trade flows and large tariffs liberalization. Among NAFTA members, Mexico was the country with the highest tariffs and experienced the largest cut. However, thanks to the high tariffs imposed by the US against the rest of the world, Mexico also enjoyed large preferential margins. 1 It is also important to consider that the peso devaluation of December 1994, which lead to large drops in output during the period 1994-1995, might also play a role in the labor market reallocation. This paper is related to three strands of literature. Firstly, we contribute to the literature that studies the effect of international trade on labour market outcomes, in particular to the small group of papers that investigate the role of informality. Exploiting sector-specific changes in tariffs, Aleman-Castilla (2006) and McCaig and Pavcnik (2014) analyse the role of trade liberalization on informality rates in Mexico and Vietman. We complement those papers by investigating whether changes in informality rates induced by trade liberalization differ between gender and across occupations. Second, our work also relates to the local labour market approach, which analyses the relationship between regional trade exposure and labour market outcomes. Chiquiar (2008) studies the effect of trade liberalization on wages across Mexican regions. Topalova (2010) evaluates the effect of trade liberalization on regional poverty using variation in sectoral composition across Indian districts. Autor et al. (2013) study the effect of rising imports from China on employment outcomes 1 See Kose et al. (2004) for a review of the literature and a discussion of this issue. 2

in U.S. local labour markets. In the literature on local labour markets and international trade, the link between informality and liberalization remains unexplored. This paper intends to fill this gap and investigate whether local exposure to changes in tariffs affects gender sorting across formal and informal employment. Finally, our approach relates to a small but growing group of papers that emphasizes the role of international trade for gender differences. This group of papers has investigated how trade liberalization affects gender participation gaps and gender wage gaps looking at different channels. Trade liberalization can contribute to a reduction in the gender gap due to taste-based discrimination through a competition effect. 2 Other papers stress the importance of male and female s characteristics, which can be specific to certain activities. This in turn can account for the different effects of international trade on their labour market outcomes (see Galor and Weil (1998), Black et al. (2007), and Sauré and Zoabi (2014) among others). (1996), Fontana et al. The literature on trade liberalization and gender inequality has focused on wage employment, and has overlooked the importance of informality as well as the role of self-employment, while self-employment represents a key source of income in developping countries and recent evidence has shown that self-employment adjusts to changes in export market opportunities (McCaig and Pavcnik (2014)). 3 This paper tries to fill this gap. Moreover, by examining gender outcomes at the municipality level, we are able to explore both the direct effect of trade liberalization on women s and men s occupational sorting in tradable sectors, as well as the total effect of trade liberalization on occupational sorting across all sectors of the local economy. To describe the evolution of occupational sorting for men and women in Mexico, we use micro data from the ENEU (Encuesta Nacional de Empleo Urbano) between 1993 and 2001. To maintain consistency across years, we present the main results restricting the sample to 55 municipalities that have been consistently surveyed. We group the working individuals into three categories: formal and informal wage workers and self-employed individuals. An individual is formally self-employed if she/he has registered her/his own business. In Mexico, women are more likely to be wage workers than men. Formal wage employment represents between 50 and 60 percent of female employment, and be- 2 The taste-based discrimination theory, along the lines of Becker (1971), suggests that gender prejudices result in higher gender gaps in concentrated industries because employers can afford to forgo profits and favor men. As trade liberalization increases competition and forces firms in concentrated industries to cut their costs, it reduces gender gaps due to discrimination in industries that enjoyed rents prior liberalization episode. See for example Black and Brainerd (2004) and Ben Yahmed (2012). 3 For example, Fox and Sohnesen (2012) show that even in times of high economic growth, household enterprises, which include also self-employed workers, generate the majority on new non-farm jobs in many Sub-Saharan countries 3

tween 48 and 55 percent of male employment. The share of informal employment is also higher among females (around 26 percent) compared to males (around 20 percent). Men, however, are more likely to be entrepreneurs. Descriptive statistics show that self-employed individuals run very small businesses. Men are more likely than women to register their business. We combine the employment data with Mexican and U.S. tariffs at the 4-digit industry level. To investigate reallocation effects between-industries within different Mexican regions, we construct an employmentweighted tariff variable that varies at regional level. The combined datasets enable us to analyse trade related dynamics in occupational sorting for men and women. Before NAFTA, labour has been reallocated from formal wage employment into informal wage and self-employment. This reallocation was more pronounced among female workers. After the devaluation event, we observe the reverse, as labour is reallocated from self-employment and informal wage employment into formal wage employment. This paper tries to quantify the role of trade integration in shaping these employment reallocations. We decompose the change in formal employment share into between and withinindustry employment shifts among the male and female working population. We find that at the 1-digit between-industry shifts, between services and manufacturing, do explain part of labour reallocation across formal and informal jobs, both for men and women. That being said, within-industry changes explain the larger part of total changes in occupational shares. Even at the 4-digit level of disaggregation, withinindustry changes continue to explain a significant part of total changes in occupational shares. These aggregate trends will motivate a two-fold empirical strategy which consists in examining how trade liberalization affects labour reallocation across formal and informal jobs first within sectors and second between sectors. We use sector specific tariffs and construct municipality tariffs to quantify the effect of trade liberalization on occupational sorting. Controlling for industry timeinvariant characteristics, we find that a fall in sector specific Mexican import tariffs increases the probability to become a formal wage worker relative to an informal worker or self-employed individual; this effect is stronger for women. To exploit regional differences in industrial composition we construct municipality exposure to tariff change. Controlling for micro-regions fixed-effects, we find that regions more exposed to falls in import tariffs due to their industrial composition experience more job reallocation from informal wage employment into formal wage employment. The paper is structured as follows. Section 2 provides a description of the data used in this study. Section 3 presents preliminary statistics. Section 4 describes the empirical strategy. Estimation results and robustness checks are discussed in Section 4

5. Section 6 concludes. 2 Data The data used in this study come from the Encuesta Nacional de Empleo Urban (ENEU), which is a quarterly labor force survey conducted by the Mexican National Statistic Institute (INEGI). 4 ENEU is a rotating panel, in which each individual can be surveyed for a maximum of 5 consecutive quarters. Each quarter a new cohort of individuals is selected, and one fifth of existing individuals leave the sample. However, particularly, for the earlier years of the survey, the panel data would not allow a gender disaggregated analysis, due to the restricted number of observations. The survey, which is representative of cities with over 100.000 inhabitants, covers only the urban areas and the primary sampling units are municipalities. In the period under analysis, 1989-2001, the sample size of the survey has constantly increased since cities were added every year. Thus, to maintain consistency across years, we present the main results focusing only on the subset of municipalities surveyed in the period 1989-2001. This implies restricting the sample to 55 municipalities that have been consistently surveyed. 5 Table 1: Data description Dependent variables Number per year Observations per year Observations mean max min mean max min 1989-2001 Industry (4-digit) 371 399 352 7,289 38,482 0 3,321,423 Export tariff 84 99 67 1,308 5,102 0 853,385 Import tariff 141 179 96 1,222 5,102 0 826,042 Municipality 55 55 55 9,044 18,318 16 3,321,423 Notes: ENEU for the period 1989-2001. The first line displays statistics on the number of industries with non-missing observations across years; the mean, minimum and maximum worker-industry-year observations across industry-year cells; and the total number of worker-industry-year over the sample period. The fourth line does the same for municipalities. We restrict our sample to employed individuals between the ages of 16 and 60. The ENEU survey provides detailed labour market information and other individual and household characteristics. Importantly for our study, it allows to identify informal employment and self-employment, employment margins that are not observed in conventional firm-level data. 4 This survey has been used by several authors, including for example Robertson (2005), Verhoogen, E.A. (2008) and Bosch and Manacorda (2010). 5 A list of these municipalities is contained in the appendix. 5

We focus on three different occupations: formal and informal wage workers and self employed. We define informal wage workers as those workers who do not have any health insurance or social security coverage that should be provided through the employment contract. Firms in the ENEU survey are very small: the largest firm has 11 employee and about 98 percent of business have no employee. While a comparison of hourly earnings could provide a picture on the profitability of each occupation, the ENEU unfortunately reports only monthly equivalent earnings from which is not possible to compute daily wages. 6 We supplement the ENEU data with the information on Mexican tariffs imposed on imports from U.S. and U.S. MFN (Most Favoured Nation) tariffs. Mexican import tariff data follow CMAP classification as explained in Iacovone and Javorcik (2010). 7 U.S. tariffs are taken from John Romalis. The tariff information is used at the industry level and also to compute a local tariff exposure at the municipality level. The latter is constructed as an employment weighted average of ad valorem industry tariffs. 3 Preliminary Evidence 3.1 The labour force population The gender gap in labour market participation rate has shrunk over the 1990s. Figure 1 shows that during that period female participation rose from 38% to 46% while male participation increased by 2 percentage points. In this paper we consider how the composition of employment has evolved during that decade. In particular, we study how formal employment has changed for men and women following trade liberalization. As depicted in figure 2, female intensity among formal workers has increased overall from 38% to 40% between 1993 and 2001. This increase has been stronger in the manufacturing industries that include almost all tradable industries as it rose from less than 32% in 1993 to 37% in 2001. Over the same period, Mexican tariffs applied on U.S. goods have fallen sharply. Figure 2 shows that the average tariff applied at the Mexican border has fallen from more than 12% to around 2% on average in the early 2000s. 6 This is because the information on the number of working days is not available in the publicly available version of ENEU. 7 We thank Beata Javorcik for providing us with the Mexican tariff data. Tariff data was available originally at the 8-digit Harmonized System (HS) classification and was matched to the Mexican CMAP class classification as explained in Iacovone et al. (2015) and Iacovone and Javorcik (2010). More information on the tariff data is provided in appendix Table?. 6

Figure 1: Labour market participation by sex Source: ENEU, Mexico. In the analysis, we control for changes in the characteristics of the working population in particular in terms of skill level. Figure 3 shows the evolution of the educational level among working individuals. Compared to women, men had in the early 1990s higher shares of workers at the bottom and at the top of the education distribution, indicating that selection into the labour force followed very different patterns for men and women. However, those differences have shrunk in the 1990s and the female and male working population become more similar in terms of educational attainment. Moreover, the level of education has increased among the overall working population. The share of people with less than 7 years of education has been falling steadily during the 1990s and represents around 30% of both the female and the male labour force in 2000. This fall has been slightly more important among women. In the meantime, the share of workers with high levels of education, above 9 years, has strongly increased. It went up from 27% (respectively 20%) of the female (male) working population in 1989 to around 35% of the female and the male labour force in 2000. In figure 4, we can see that the increase in the average level of education has been pervasive for all types of occupations. 7

Figure 2: Female labour intensity and Mexican tariffs Source: ENEU, Mexico. In Figure 5 we plot the occupational shares respectively for the female and the male working population. Men are more often self-employed than women. Selfemployment accounts for 26 to 30% of the male working population, while it represents between 16% 20% of the female working population. On the other hand, the share of wage workers is higher in the female working population than in the male working population, especially for formal wage workers. Formal wage employment represents between 53 and 58% of the female working population and only 49 to 54% of the male working population. The gender employment gap in formal wage employment was bigger at the start of the period, i.e. 10 percentage points in 1989. It was reduced to 5 percentage points in 2000. Women are also more often in informal wage employment, however we dot not observe here a convergence in the male and female informality share.the gender gap in informal wage employment has increased from 3 percentage points in 1989 to 6 percentage points in 2001. This is due to a much more rapid increase in informal employment among women in the first part of the 1990s, while the fall in informality in the second part of the 1990s has benefited both sex groups. Table 2 shows that informality is common among both blue and white collar workers as well as among both unskilled and skilled workers. Women are more often in white-collar positions. A very high share of informal female wage workers and female self-employed are in low-skilled white-collar jobs. Men tend to be more evenly represented in blue and white-collar jobs. Table 8 in the appendix gives the precise repartition of men and women across occupation titles for formal and informal wage 8

Figure 3: Evolution of educational level among employed individuals by sex Source: ENEU, Mexico. employment and self-employment in 1992. More descriptive statistics are provided in the appendix. The repartition of men and women across industries is given in table 9. Table 10 shows descriptive statistics on the final sample used in the empirical analysis. 9

Figure 4: Evolution of the average educational level by occupation and sex Source: ENEU, Mexico. Table 2: Repartition across blue and white-collar occupations in 1992 Formal wage workers Informal wage workers Self-employment (1) (2) (3) (4) (5) (6) Men Women Men Women Men Women High-skill White-collar 26.99 38.40 17.91 21.85 17.12 9.89 Low-skill White-collar 24.70 38.20 25.61 68.48 25.60 61.98 High-skill Blue-collar 26.76 11.40 26.06 5.46 36.50 13.72 Low-skill Blue-collar 21.55 12.00 30.42 4.20 20.78 14.41 Total 100.00 100.00 100.00 100.00 100.00 100.00 10

Figure 5: Occupational status by sex Source: ENEU, Mexico. 11

3.2 Between and Within Decomposition In what follows, we explore how formal employment shares have evolved across industries separately for men and women. We decompose the change in formal employment share into two components. The between-industry component reflects labour reallocation between industries with different formal employment shares. The within-industry component reflects labour reallocation across formal and informal jobs within industries. Both types of reallocation can be influenced by trade, but through different mechanisms. In a comparative advantage framework, trade integration increases production in comparative advantage industries and contributes to the decline of others, thereby inducing between-industry reallocation of labour. If formal labour intensities differ in comparative advantage industries compared to other industries, the overall formal labour share will change with trade. If the comparative advantage industries have different female labour intensity compared to other industries, trade will affect formal employment shares differently for men and women. In a framework with intra-industry trade and heterogeneous firms, trade integration generates within-industry changes through firm selection and technology upgrading. Assuming that trade-oriented firms are more intensive in formal labour, trade liberalization is expected to increase the formal employment shares within-industry. Moreover, gender-biased technological change, or changes in gender discrimination because of tougher foreign competition would shift the gender composition of employment within industries. A decomposition of the changes in formal employment shares for men and women into changes that occur within and between industries can thus provide useful indication about the sources behind the evolution of the gender gap in formal employment. We implement the following decomposition for each of the occupations under study, among the female and male working population separately: Occ gt = i Occ git E gi + i E git Occ gi (1) for i = 1,..., N industries and Occ = {Formal jobs; Infrmal jobs}. Occ gt denotes the share of a given occupation in total employment of group g = {f, m} where f denotes female and m denotes male. Occ gt = Occ gt Occ gt 1 is the change in occupational share over the period. Occ git is the share of a given occupation in industry i s employment of group g while E gi is industry i s employment of group g. Equation (1) decomposes the aggregate change in occupational share into within-industry changes (the first term on the right hand side), and between-industry changes (the second 12

term). The within-industry component reports the change in the aggregate share of the occupation due to variations in occupational shares within each industry, holding constant the industry employment share. The between-industry component reflects the change in the aggregate share of the occupation due to changes in employment shares between industries, holding constant the occupational shares within industries. We decompose the changes in occupational shares for the overall period following the creation of the NAFTA 1994-2001 and for two sub-periods, 1994-1998 and 1998-2001. Table 3 shows how formal wage employment has evolved relative to informal wage employment in the overall economy (including both tradable and non-tradable sectors). It displays the results of the decomposition using the 1-digit level industry classification. Including all sectors allows us to account for reallocation across broadly defined industries, in particular between tradable and non-tradable sectors. Table 3 shows that formal wage employment has increased in the overall economy compared to informal employment. The increase in formal employment relative to informal employment took place at the end of the 90s. Between 1994 and 1998, informal employment slightly increased, probably in the aftermath of the peso devaluation and financial crisis. The end of 90s is characterized by a formalization of employment. The increase in formal employment and fall in informal employment is only due to within-industry changes. However, and contrary to within-industry changes, employment changes between 1-digit industries contributed to reallocation of labour from formal to informal jobs. It it thus important to explore the effect of trade liberalization on informality on the overall economy including non-tradable industries, as the effect differ from the within industry effect. Table 4 shows the result of the decomposition on the sub-sample of manufacturing industries. Between 1994 and 2001, formal employment has increased relatively to informal wage employment and self-employment. At the 3-digit level of sector disaggregation, within-industry change explain most of the total changes in formal employment shares. Differently than for the decomposition on the overall economy, the decomposition on the manufacturing sectors shows that both within-industry and between-industry changes have contributed to a formalization of jobs in the manufacturing sectors, and this all along the second part of the 90s. Gender differences in the magnitude of the changes distinguishes the first period 1994-1998 from the second period. Between 1998 and 2000, the formal employment share has increased by 16 percentage points for women and by only 5 percentage points for men. The formalization of jobs in the manufacturing industries is thus much more pronounced 13

Table 3: Decomposition of changes in occupational shares into between-industry and within-industry shifts across 1-digit industries. Formal wage employment Informal wage employment Self-employment Total Within Between Total Within Between 1994-2001 Female.03.021.01 -.03 -.021 -.01 Male.031.017.013 -.031 -.017 -.013 1994-1998 Female -.006 -.013.007.006.013 -.007 Male -.003 -.013.01.003.013 -.01 1998-2001 Female.009.011 -.002 -.009 -.011.002 Male.007.009 -.003 -.007 -.009.003 Source: Author s calculation based on the ENEU. among the female working population. Again this is mostly due to reallocation from informal to formal jobs within industries. Changes within 1-digit and even 3-digit industries explain the bigger part of the overall changes in formal employment shares. However, between-industry changes also matter, especially at the 1-digit level, which motivates the analysis of the role of trade liberalization in reallocation across tradable and non-tradable industries.these results support the choice of two specifications to capture the effect of trade integration on occupational sorting across gender both within and between industries. First, using tariff at the sectoral level to explore whether trade liberalization has affected men s and women s occupational sorting within disaggregated tradable industries. Second, exploiting changes in regional exposure to tariff changes in order to investigate whether trade liberalization affects men s and women s occupational sorting through reallocation between industries including non-tradable sectors. The following section explains how those two tariff measures are computed. 14

Table 4: Decomposition of changes in occupational shares into between-industry and within-industry shifts across 3-digit manufacturing industries. Formal wage employment Informal wage employment Self-employment Total Within Between Total Within Between 1994-2001 Female.051.045.005 -.051 -.045 -.006 Male.043.027.016 -.043 -.027 -.017 1994-1998 Female.035.031.001 -.035 -.031 -.004 Male.038.022.015 -.038 -.022 -.016 1998-2001 Female.016.014.002 -.016 -.014 -.002 Male.005.004.001 -.005 -.004 -.001 Source: Author s calculation based on the ENEU. 15

4 Empirical Strategy This section describes the different tariff measures used and then presents the empirical methodology. Since we focus on three types of occupational category, formal and informal wage workers and self employed, our definition of working population includes individual active in formal and informal wage sector and self employment. 4.1 Measures of Trade Liberalization First, we test the within industry impact of trade liberalization on male and female employment in formal and informal jobs. We use ad-valorem industry tariffs, τ st, which vary across 4-digit manufacturing sectors s and time t. The tariff measure enables us to exploit within variations in tradable industries across time to study how trade policy affects gender sorting across occupations. Then, we analyze the effect of exposure to international trade on local labor market across industries within regions (municipalities). To measure a municipality s exposure to trade, we compute an employment-weighted average of ad-valorem industry tariffs of the manufacturing industries active in that municipality, T ariff mt. The weight is given by the number of workers in that industry as a share of all workers in the district (at the beginning of the period). This municipality tariff variable is thus computed as follows: τ mt = ST r s ST r Empl ms τ st s Empl ms (2) where s indicates the industry, m the municipality, and t the year. We are then able to exploit regional differences in industrial composition and different degree of liberalization across industries to identify the effect of trade liberalization on occupational choices within regions across industries. Finally, to better understand how aggregate outcomes, such as labor force participation rates, react to trade, we also extend the analysis to non-tradable sectors. In this case the municipality tariff is constructed assigning a zero tariff to non-traded industries, which implies that here the employment-weighted average of ad-valorem industry tariffs includes all industries active in that municipality. Therefore, the municipality tariff will be sensitive to the share of people working in non-tradable industries. τall mt = Sall s Sall Empl ms τ st s Empl ms (3) where s S all indicates an employment-weighted average which accounts for all 16

industries. The municipality tariff in equation (3) is sensitive to the employment share of sectors producing non-tradable goods. This could introduce confounding factors in the empirical specification if regions with higher employment shares in non-tradable sectors have different evolutions of gender sorting across occupations for reasons unrelated to trade liberalization. This problem will be corrected when using equation (2) that instead uses only employment in traded production sectors to weight the tariff measure (note that then the denominator is smaller than the one in (3) and that the variable does not depend on the employment composition across tradable vs. non-tradable sectors). 4.2 Empirical Implementation Specification at Individual Level To assess the importance of increased trade openness on male and female employment in formal and informal jobs, we first look at the effect of sectoral changes in ad valorem tariffs on the formality rates within sectors. Since NAFTA implied a reduction in import as well as export tariffs, we explicitly account for them in our empirical specification. Thus we estimate the following linear probability model: F ijsmt = αx ijsmt + β 1 τ Mex st β 3 τ US st + β 2 τ Mex st female i + + β 4 τ US st female i + µ s + u m + v t + ɛ ijmt (4) where the subscript i denotes the ith individual, subscripts j denotes the alternative, X ijsmt is a vector of specific individual characteristics (age, age squared and education). τ Mex st and τ US st are the Mexican import and export tariff respectively. F emale i female i and is a dummy variable equal to one if the individual is a female. τst Mex female i are interaction terms between the average import and export tariffs τ US st and the female dummy. The coefficients of these interaction terms, β 2 and β 4, capture how the Mexican female working population has been affected by trade liberalization. The sum between the coefficients β 1 and β 2 gives the average effect of the reduction in Mexican import tariff on female. While, the sum between the coefficients β 3 and β 4 gives the average effect of the reduction in US import tariff on female. µ s, u m and v t are a full set of 4-digit sector, municipality, and year fixed effects. The theoretical framework predicts a negative and significant effect for our coefficients of interest, β 1 and β 2 : a decline in import and export tariffs should increase the size of the formal sector. Moreover, we expect this effect to be stronger for female (β 2 and β 4 negative and significant). This specification should not suffer from endogeneity of the trade tariff because changes at the individual level are unlikely to 17

exert an impact on industry aggregates. Equation (4) is estimated on the sample of the working population in tradable industries. To account for the regional exposure to trade liberalization on labour reallocation across tradable sectors, we use a specification similar to (4) where the industry level tariffs are replaced with regional level tariffs which vary across municipalities and years. The probability of the ith individual to be working in occupation j is then given by: F ijmt = αx ijmt + β 1 τ Mex mt β 3 τ US mt + β 2 τ Mex mt female i + + β 4 τ US mt female i + u m + v t + ɛ ijmt (5) Trade exposure is then identified by using variation in tariffs within municipalities across years. Equation (5) is estimated on the sample of all working population in tradable industries. Then to account for different between industry effect in tradable from tradable industries, we estimate equation (5) on the full set industries. Specification at Regional Level In our theoretical set up, Mexico has a comparative advantage in female labor intensive sectors (due to technology or factor aboundance). Therefore, in these sectors we should observe an increase in relative employment of female workers associated with tariff reductions. In this section, we relate changes in labor market outcomes from 1993 to 2001 across Mexican local labor markets to changes in exposure to the US market. To analyze the relationship between Mexico manufacturing employment and trade liberalization, we fit a model of the following form: F emaleshare st = αx st + β 1 τ Mex st + β 2 τ US st + µ s + v t + ɛ st (6) where F emaleshare st is the share of female to male workers in the working population in each manufacturing sector, s, across time, t: F emaleshare st = i F emale ist i F emalew orkp op ist i Male ist i MaleW orkp op ist (7) Equation (7) will be computed for both formal and informal workers in the working population. To test the between sector reallocation in each municipality, we also estimate the 18

following equation: F emaleshare mt = αx mt + β 1 τall Mex mt + β 2 τall Mex mt female i + β 3 τall US mt + β 4 τall US mt female i + u m + v t + ɛ mt (8) where F emaleshare mt is the share of female in formal jobs in the female population relative to the share of male formal workers in the male working population in each municipality, m, across time, t: F emaleshare mt = i F emale F ormal imt i F emale W orkp op imt i Male F ormal imt i Male W orkp op imt (9) To account for unemployment issues, we also perform a similar exercise adjusting our share measure to account for unemployment in the working population. 5 Empirical Results 5.1 NAFTA and Gender Sorting across Occupations withinindustries We begin by examining the effect of sector specific tariffs on the probability to hold a formal job and estimate equation (4). The estimates provide an indication on the effect of tariffs on formalization within industry. Table 5 reports the results for the working population in tradable sectors in columns (1) and (2), for the sub-samples of the working population in tradable sectors living in non-border regions in columns (3) and (4), and for workers in low-skill occupations in columns (5) and (6) 8. Trade liberalization leads to a formalization of job within 4-digit manufacturing industries. The reallocation of jobs from informal into formal wage employment benefits both men and women but seems to be stronger for women as the results in columns (1) and (3) suggest. This formalization of employment results from both a fall of Mexican and U.S. tariffs. Controlling for both Mexican and U.S. tariffs (in columns (2) (4) and (6)), the effect of trade liberalization on the probability to be in formal wage employment is overall similar for men and women. A notable exception is workers in low-skill occupations. Among those workers, only men have a higher probability to be in formal employment following the U.S. trade liberalization. On the contrary, women 8 Low-skill occupations include agricultural workers, machinery workers, trade workers, elementary workers, drivers, street and related workers, and domestic workers. 19

in low-skill occupations are more likely to be in informal employment in industries that experience larger fall in U.S. tariff. Table 5: Formal Wage Employment and Within-industry Tariff Changes (1) (2) (3) (4) (5) (6) all all Non-border Non-border Low-skill Low-skill Mex Tariff -0.405*** -0.299*** -0.381*** -0.286*** -0.208*** -0.204*** (0.097) (0.079) (0.100) (0.085) (0.070) (0.068) Mex Tariff female -0.166* -0.107-0.166* -0.114 0.046-0.037 (0.093) (0.091) (0.100) (0.098) (0.063) (0.063) US Tariff -0.386** -0.380** -0.245* (0.180) (0.179) (0.148) US Tariff female -0.185-0.129 0.567** (0.359) (0.355) (0.227) age 0.016*** 0.016*** 0.020*** 0.020*** 0.024*** 0.024*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) age 2-0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) school 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) female 0.010 0.009 0.003 0.002-0.001-0.004 (0.009) (0.009) (0.011) (0.011) (0.005) (0.005) Constant 0.915 0.912 0.738*** 0.736*** 0.303 0.312 (.) (39.783) (0.034) (0.032) (.) (253.432) Observations 469,842 469,842 334,573 334,573 266,218 266,218 R-squared 0.288 0.288 0.245 0.245 0.216 0.216 Year FE Yes Yes Yes Yes Yes Yes YearxRegion FE No No No No No No Municipality Yes Yes Yes Yes Yes Yes 4-digit sector FE Yes Yes Yes Yes Yes Yes Notes: Linear probability model estimation. Clustered s.d. at the sector x year level in parentheses. The dependent variable equals 1 if the individual is in formal wage employment and 0 if she is in informal wage employment or self-employment. Only manufacturing sectors. 5.2 The NAFTA and Gender Sorting across Occupations at the Local Labour Market Level Table 6 shows the estimation results for equation (5) where we adopt a local labour market approach. In this specification, we regress an indicator for formal wage employment on municipality exposure to U.S. tariffs and municipality exposure to Mexican tariffs that are computed as explained in section 4.1. The two local tariff measures 20

vary across time and municipalities but not across sectors. We additionally control individual characteristics and for municipality and year fixed-effects. This specification allows us to explore the effects of trade liberalization on reallocation between formal and informal employment across industries. Table 6 reports results on the sample of workers in tradable industries in order to assess whether trade liberalization changes the size of the formal segment of the labour market through labour reallocation between tradable industries. Table 7 reports results on the sample of all workers, including tradable and non-tradable sectors, in order to assess whether trade liberalization changes the relative size of the formal segment of the labour market through labour reallocation between the tradable and the non-tradable sectors. Table 6 reports the estimates of local exposure to tariffs on the probability to be in formal wage employment for workers in manufacturing sectors only. Local exposure to cuts in Mexican tariff affects the probability of being formally employed in the manufacturing sector for women only. First, the negative coefficient associated with local exposure to Mexican tariffs interacted with the female dummy (except in border regions) shows that a fall in regional exposure to import tariffs increases the probability to be an formal employee for women. However, columns 2 and 6 show that on the contrary women working in municipalities more exposed to cuts in U.S. tariff experienced an increase in the probability to be in informal wage employment or self-employed compared to be a formal employment. The results are reverse for women living in municipalities situated at the border with the United-States or among women in high-skill occupations. These indicates that direction of trade liberalization matters. It is thus important in a next step to take into account actual trade flows and to distinguish comparative advantage or export-oriented from import competing sectors. Moreover, the regional heterogeneity in terms of industry mix between nonborder and border regions is also important. The specificity of the maquiladoras sector, mostly located at the border, is also a dimension that should be integrated to the analysis. Table 7 reports the estimates of local exposure to tariffs on the probability to be in formal wage employment for workers in all sectors. The significant effect of the local exposure to Mexican tariff indicates that trade liberalization affects the probability of being formally employed beyond the tradable sectors. First, and differently from the results on within-industry reallocation, the positive coefficient associated with local exposure to Mexican tariffs (column 1 to 5) shows that a fall in regional exposure to import tariffs increases the probability to be an informal employee for men. Women also have a higher probability to be informally employed in municipalities that are exposed to large fall in Mexican tariff but less so than men as indicated by the neg- 21

Table 6: Formal Wage Employment and Changes in Municipality Exposure to Tariffs Manufacturing sectors only (1) (2) (3) (4) (5) (6) all all border border High-skill High-skill Muni Mex Tar -0.089-0.131 0.501-0.353 0.393 0.097 (0.181) (0.229) (0.867) (1.012) (0.241) (0.291) Muni Mex Tar female -0.521*** -0.943*** 0.100 0.740** -0.817*** -1.172*** (0.134) (0.186) (0.114) (0.349) (0.157) (0.199) Muni US Tar -0.119 3.147 0.789 (0.345) (2.546) (0.487) Muni US Tar female 1.602*** -4.756* 1.447*** (0.408) (2.607) (0.515) age 0.019*** 0.019*** 0.010*** 0.010*** 0.008*** 0.008*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) age 2-0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) school 0.009*** 0.008*** 0.006*** 0.006*** 0.026*** 0.025*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) female 0.045*** 0.044*** 0.075*** 0.073*** 0.017 0.015 (0.008) (0.008) (0.007) (0.007) (0.013) (0.013) Constant 0.504*** 0.520*** 0.566*** 0.631*** 0.406*** 0.432*** (0.038) (0.040) (0.121) (0.125) (0.054) (0.053) Observations 493,674 493,674 139,918 139,918 210,417 210,417 R-squared 0.124 0.124 0.092 0.092 0.179 0.179 Year FE Yes Yes Yes Yes Yes Yes YearxRegion FE No No No No No No Municipality Yes Yes Yes Yes Yes Yes 4-digit sector FE No No No No No No Notes: LPM estimation. Clustered s.d. at the sectorxyear level in parentheses Dependent variable equals to 1 if in formal wage employment and to 0 if in informal wage employment or informal self-employement. Only manufacturing sectors. 22

ative coefficient associated with the interaction between tariff and a female dummy. Second columns 4 and 6 show that municipalities more exposed to cuts in U.S. tariff experienced an increase in the probability to be in informal wage employment or selfemployed compared to be a formal employment for women in municipalities situated at the border with the United-States or among women in high-skill occupations. 23

Table 7: Formal Wage Employment and Changes in Municipality Exposure to Tariffs (1) (2) (3) (4) (5) (6) all all border border High-skill High-skill Muni Mex Tar 0.519*** 0.398** 0.482*** 0.523*** 0.415*** 0.290 (0.148) (0.154) (0.155) (0.142) (0.160) (0.181) Muni Mex Tar female -0.164** -0.197** 0.033-0.198*** -0.034-0.267** (0.077) (0.096) (0.062) (0.067) (0.081) (0.105) Muni US Tar 0.349-0.177 0.283 (0.295) (0.309) (0.341) Muni US Tar female 0.137 0.957*** 0.971*** (0.220) (0.145) (0.278) age 0.013*** 0.013*** 0.018*** 0.018*** 0.013*** 0.013*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) age 2-0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) school 0.021*** 0.021*** 0.023*** 0.023*** 0.023*** 0.023*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) female 0.045*** 0.045*** 0.010* 0.008 0.134*** 0.133*** (0.007) (0.007) (0.006) (0.006) (0.007) (0.007) Constant 0.156*** 0.165*** 0.051* 0.052* 0.087** 0.100*** (0.035) (0.035) (0.030) (0.030) (0.037) (0.038) Observations 2,030,971 2,030,971 1,595,867 1,595,867 824,725 824,725 R-squared 0.069 0.069 0.066 0.066 0.087 0.087 Year FE Yes Yes Yes Yes Yes Yes YearxRegion FE No No No No No No Municipality Yes Yes Yes Yes Yes Yes 4-digit sector FE No No No No No No Notes: Linear probability model. Clustered s.d. at the municipality x year level in parentheses. The dependent variable equals 1 if in formal wage employment and 0 if she is in informal wage employment or self-employement. Sample of all individuals in both tradable and non-tradable sectors. 24

References Aguayo-Tellez, E., Airola, J., Juhn, C., 2010. Did Trade Liberalization Help Women? The Case of Mexico in the 1990s. Research in Labor Economics, 38,. Aleman-Castilla, B., 2006. The Effect of Trade Liberalization on Informality and Wages: Evidence from Mexico. Centre for Economic Performance (CEP), Discussion Paper 763. David Autor, David Dorn, Gordon Hanson, 2013. The China Syndrome: Local Labor Market Effects of Import Competition in the United States. American Economic Review, 103(6), 2121 2168. Ben Yahmed, S., 2012. Gender Wage Discrimination and Trade Openness. Prejudiced employers in an open industry. Aix-Marseille School of Economics Working paper series, 2012-33. Batini, Nicoletta and Levine, Paul and Kim, Young-Bae and Lotti, Emanuela, 2010 Informal labour and credit markets: a survey, IMF Working Papers, 1 41. Black, S. E., Brainerd, E., 2004. Importing equality? The impact of globalization on gender discrimination. Industrial and Labor Relations Review, 57(4), 540 559. Bosch, M., Manacorda, M., 2010. Minimum Wages and Earnings Inequality in Urban Mexico. American Economic Journal: Applied Economics, American Economic Association, 2(4), 128 49. Bustos, P., 2011. Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of MERCOSUR on Argentinian Firms. American Economic Review, 101(1), 304 340. Chiquiar, D., 2008. Globalization, regional wage differentials and the Stolper- Samuelson theorem: evidence from Mexico. Journal of International Economics, 74(1), 70 93. Do, Q., Levchenko, A., Raddatz, C., 2011. Comparative Advantage, International Trade, and Fertility. Journal of Development Economics, 119 (March 2016), 48-66.. 119(1), 48 66. Duflo, E., 2012. Women Empowerment and Economic Development. Journal of Economic Literature, 50(4), 1051 79. 25

Feenstra, Robert C. and John Romalis and Peter K. Schott 2002. U.S. Imports, Exports, and Tariff Data, 1989-2001 NBER Working Paper No. 9387 Fontana, M., Joekes, S.P., Masika, R., 1998. Global trade expansion and liberalisation: gender issues and impacts, BRIDGE Institute of Development Studies Brighton, Report No 42. Fox, L., Sohnesen, T.P., 2012. Household Enterprises in Sub-Saharan Africa. Why They Matter for Growth, Jobs, and Livelihoods. World Bank Policy Research WP No.6184. Gaddis, I., Pieters, J., 2014. The Gendered Labor Market Impacts of Trade Liberalization: Evidence from Brazil. World Bank Policy Research WP No.7095. Galor, O., Weil, D.N., 1996. The Gender Gap, Fertility, and Growth. American Economic Review, 86(3) 374 387. Iacovone, L., Javorcik, B., Keller, W.,Tybout, J., 2015. Supplier Responses to Walmart s Invasion in Mexico. Journal of International Economics, 95(1), 1 15. Iacovone, L., Javorcik, B., 2010. Multi-product Exporters: Product Churning, Uncertainty and Export Discoveries. Economic Journal, 120(544). Juhn, C., Ujhelyi, G., Villegas-Sanchez, C., 2013. Trade Liberalization and Gender Inequality. American Economic Review, Papers and Proceedings, 3(103). Juhn, C., Ujhelyi, G., Villegas-Sanchez, C., 2014. Men, Women, and Machines: The Impact of Trade on Gender Inequality. Journal of Development Economics, 106,. McCaig, B., Pavcnik, N., 2014. Export Markets and Labor Allocation in a Low- Income Country. NBER Working Paper No. 20455. Sauré, P., Zoabi, H., 2014. International Trade, the Gender Wage Gap and Female Labor Force Participation. Journal of Development Economics, 111, 17 33. Topalova, P., 2010. Factor immobility and regional impacts of trade liberalization: Evidence on poverty from India. American Economic Journal: Applied Economics, American Economic Association, 2(4), 1 41. Verhoogen, E., 2008. Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing Sector. Quarterly Journal of Economics, 123(2),. 26

Appendix Table 8: Repartition across occupations in 1992 (1) (2) (3) (4) (5) (6) Formal wage workers Informal wage workers Self-employment Men Women Men Women Men Women Agriculture workers 0.24 0.01 1.87 0.15 2.39 0.26 Army 5.30 0.27 4.23 0.11 2.06 0.04 Arts and sports professionals 0.64 0.35 1.29 0.46 1.65 0.80 Clerks 7.62 22.04 4.47 11.88 0.23 0.21 Craft workers 20.15 8.54 23.43 5.02 34.67 13.52 Domestic workers 0.28 0.46 1.58 34.12 0.58 15.39 Drivers 5.17 0.02 4.24 0.02 5.81 0.18 Education professionals 3.36 11.44 2.10 6.33 0.29 1.57 Elementary workers 5.22 2.75 14.52 1.63 0.36 0.06 Legislators and Executive Managers 3.14 1.37 2.02 1.95 4.59 1.45 Machinary workers 9.73 9.14 7.65 5.72 0.97 0.37 Managers 7.73 10.92 4.17 10.49 0.22 0.18 Personnal Service workers 7.70 7.04 8.73 2.86 5.92 14.57 Professionnals 5.39 4.77 3.61 0.45 7.09 4.77 Street sales and services workers 0.02 0.01 0.83 0.43 10.82 13.53 Supervisors and controllers 5.19 2.82 1.52 5.57 1.08 0.20 Technicians and associate professionals 5.29 9.44 3.97 11.91 2.92 1.12 Trade workers 7.80 8.55 9.74 18.35 31.79 Total 100.00 100.00 100.00 100.00 100.00 100.00 27