The World Bank Economic Review, 37(2), 2023, 305–330 https://doi.org10.1093/wber/lhad009 Article Trade Shocks, Population Growth, and Migration Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Sofía Fernández Guerrico Abstract This paper examines the effect of trade-induced changes in Mexican labor demand on population growth and migration responses at the local level. It exploits cross-municipality variation in exposure to a change in trade policy between the United States and China that eliminated potential tariff increases on Chinese imports, negatively affecting Mexican manufacturing exports to the United States. Municipalities more exposed to the policy change, via their industry structure, experienced greater employment loss. In the five years following the change in trade policy, more exposed municipalities experience increased population growth, driven by declines in out-migration. Conversely, 6 to 10 years after the change in trade policy, exposure to increased trade competition is associated with decreased population growth, driven by declines in in-migration and return migration rates, and increased out-migration. The sluggish regional adjustment is consistent with high moving costs and transitions across sectors in the short term. JEL classification: F16, J23, O12, R12, R23 Keywords: trade competition, job displacement, population growth, internal migration 1. Introduction Why, when, and where do individuals decide to migrate? Beyond its intrinsic relevance, the answer to this question has important implications for the estimates of several socioeconomic outcomes. Selective migration in response to an economic shock changes the composition of local labor markets. Conse- quently, estimated impacts on average socioeconomic outcomes might reflect both a direct treatment effect and a change in outcomes driven by changes in the labor-market composition (Greenland, Lopresti, and McHenry 2019; Arthi, Beach, and Hanlon 2022). Accounting for such compositional effects represents an empirical challenge, especially when individual panel data is not available. This paper studies how trade-induced changes in Mexican labor demand affect population growth and migration flows at the local level. I exploit cross-municipality variation in exposure to a change in trade policy between the United States and China that negatively affected Mexican manufacturing exports to the US market. I find that, in the five years following the change in trade policy, more exposed municipalities experience increased population growth, driven by declines in out-migration. These results are not driven by return migration from the United States, which also relatively decreases in more exposed areas. Conversely, 6 to 10 years after the plausibly exogenous change in trade policy, exposure to increased Sofía Fernández Guerrico is a Postdoctoral Fellow at the Department of Applied Economics (DULBEA), Université Libre de Bruxelles (ULB), Belgium; her email address is Sofia.Fernandez.Guerrico@ulb.be. A supplementary online appendix is available with this article at The World Bank Economic Review website. © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 306 Fernández Guerrico trade competition is associated with decreased population growth, driven by declines in in-migration and return migration rates, and increased out-migration. Furthermore, I find heterogeneity in the effects across population groups: the migratory response is driven primarily by less educated and manufacturing sector workers. My results indicate that exposure to trade competition affected population growth via a deterioration in labor-market opportunities in the manufacturing sector, which led to income loss due to job loss and Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 lower wages. The sluggish population response is consistent with the changes in internal migration that I document. However, the initial population growth is somewhat puzzling given the consistent negative dynamic effects on manufacturing employment and income that I find. I show that transitions across sectors are a plausible mechanism behind the reversal in population growth and out-migration. I find that there are job gains in the service sector that partially offset the job losses in manufacturing in the short term. In the long term, though, increased competition is associated with declines in employment and wages across all sectors. The slack response in service-sector employment and wages, together with high moving costs, are possible channels behind the timing of the regional adjustment. My primary empirical approach focuses on a change in trade policy between the United States and China that negatively affected Mexican manufacturing exports to the United States. In October 2000, the United States granted to China Permanent Normal Trade Relations (PNTR), which eliminated po- tential tariff increases on imports from China. In the United States, Pierce and Schott (2016) link the decline in manufacturing employment after 2000 to the surge in imports of Chinese goods in the partic- ular industries affected by PNTR. Given the technological similarity between China and Mexico at the time, the increase in Chinese exports to the United States also led to a decrease in demand for Mexi- can manufacturing products (Gallagher and Porzecanski 2007; Hanson and Robertson 2008) and hence a decline in manufacturing employment opportunities in Mexico (Utar and Torres Ruiz 2013; Mendez 2015; Chiquiar, Covarrubias, and Salcedo 2017). Building on Pierce and Schott (2016, 2020), I construct a Mexican municipality-level measure of exposure to trade competition resulting from the United States granting PNTR to China, which differentially exposed regions to increased trade competition via their industry structure. Therefore, Mexican municipalities specializing in industries in which China had an initial comparative advantage were more exposed to this change in trade policy.1 The contribution of this paper is two-fold. I identify first-order effects of increased international com- petition on Mexican trade and labor market associated with a change in trade policy between the United States and China, and I document the population response to this local labor-demand shock. To the best of my knowledge, this is the first paper studying aggregate population changes as a response to trade competition in Mexico. There are, however, a vast number of studies examining US–Mexico migration (Chiquiar and Hanson 2005; Ibarraran and Lubotsky 2007; McKenzie and Rapoport 2010; Kaestner and Malamud 2014; Caballero, Cadena, and Kovak 2018, 2019;) and migration responses to income shocks (Belot and Hatton 2012; Angelucci 2015; Kleemans 2015; Quiñones 2019). The closest-related paper is on the individual decision to migrate (Majlesi and Narciso 2018), which finds that individuals living in areas with higher exposure to international competition were more likely to migrate within Mexico between 2002 and 2005.2 To provide evidence on the channels linking the population response to the trade shock, I examine the first-order effects of PNTR on Mexican manufacturing exports and local labor markets. I analyze trade 1 Chinese competition affected Mexico directly, through an increase in imports from China, and indirectly, through in- creased competition in the US market. Fernández Guerrico (2021) shows the negative effect of both import and export competition on the Mexican local labor market. See appendix S1 in the supplementary online appendix. 2 Majlesi and Narciso (2018) use data from the Mexican Family Survey; their sample covers 100 municipalities (over- sampling rural areas) whereas my sample covers 2,382 municipalities and all working-age population covered in the Mexican Economics and Population Censuses. The World Bank Economic Review 307 competition between two southern locations in a third, northern, market generated by a change in trade policy. Previous work finds a negative impact of Chinese competition in the US market on employment and plant growth within Mexican maquiladoras (Utar and Torres Ruiz 2013). Here, using cross-municipality exposure to PNTR, I show the effect of increased export competition in the entire Mexican labor market. Furthermore, using exposure to PNTR has advantages over methods that rely on supply-driven changes in China around the time of its accession to the WTO because it is based on a specific change in trade Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 policy.3 Following an approach similar to that used by Autor, Dorn, and Hanson (2013) in the US context, related work studying the impact of Chinese competition in Mexico, I use a measure regional variation in exposure to trade using changes in Mexican or US imports per worker from China (Mendez 2015; Blyde et al. 2017; Dell, Feigenberg, and Teshima 2019; Fernández Guerrico 2021). Recent research argues that weighting the local industry shares by growth rates in Chinese exports is an imperfect way of isolat- ing the variation in industries where China experienced rapid productivity gains (Goldsmith-Pinkham, Sorkin, and Swift 2020).4 In using Pierce and Schott (2016)’s measure of industry-level exposure to China receiving PNTR, the argument is not that the trade policy is random, but that the change in the trade policy is not correlated with pre-existing trends in outcomes at the local level. While this change in trade policy has been used to study changes in local labor markets in the United States, I study the local labor- market effects of PNTR in Mexico, given the importance of the United States as an export destination of manufacturing products for Mexico. More broadly, my analysis relates to the literature on heterogeneous migratory responses to local labor- market conditions based on workers’ skills (Bound and Holzer 2000; Cadena and Kovak 2016; Utar 2018; Greenland, Lopresti, and McHenry 2019; Notowidigdo 2020), the relative importance of regional mobility compared to sectoral mobility given adjustment costs (Autor et al. 2014; Dix-Carneiro and Kovak 2017; Bartik 2018), and the relative importance of out-migration compared to in-migration (Monras 2018). Finally, this paper also relates to the extensive literature examining the effect of trade liberalization on labor-market outcomes in the last two decades (Topalova 2010; McCaig 2011; Kovak 2013; Autor, Dorn, and Hanson 2013; Autor et al. 2014; Dix-Carneiro 2014; Acemoglu et al. 2016; Pierce and Schott 2016), as well as an array of socioeconomic outcomes such as education and child labor (Edmonds, Pavcnik, and Topalova 2009, 2010), marriage and fertility (Autor, Dorn, and Hanson 2015, 2019), health and mortality (Adda and Fawaz 2020; Pierce and Schott 2020; Fernández Guerrico 2021), and crime (Dix-Carneiro, Soares, and Ulyssea 2018; Dell, Feigenberg, and Teshima 2019; Khanna et al. 2019).5 Here, using a similar identification strategy, I provide an insight into the direct effects of PNTR on Mexican exports to the United States and the impact on manufacturing employment as a pos- sible mechanism that induced migratory responses. The paper proceeds as follows. In the next two sections, I describe the data and explain the identifica- tion strategy, respectively. Then, I present the results of increased trade exposure on aggregate population growth and migration responses. I also examine the first-order effects of PNTR on Mexican labor-market outcomes to provide evidence on the mechanisms linking the population response to the trade shock. The second to last section describes a series of robustness checks. Finally, the last section concludes. 3 However, this approach also presents challenges because of possible trade spillovers to third countries. For example, Mau (2017) shows that a reduction in US tariff uncertainty arising from China’s accession to WTO also positively affected China’s exports to the European Union. 4 A recent literature has advanced our understanding of the identification assumptions for shift-share designs discussing alternative approaches to recovering causal effects—assuming exposure “shares” are as good as random (Goldsmith- Pinkham, Sorkin, and Swift 2020), and assuming exposure “shocks” are as good as random (Adão, Kolesár, and Morales 2019; Borusyak, Hull, and Jaravel 2022). 5 See Pavcnik (2017) for a literature review. 308 Fernández Guerrico 2. Data This section describes the data I use to investigate the relationship between international competition, labor demand, population growth, and migration responses. 2.1. Population Growth and Migration Response across Municipalities Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Population data comes from the 2000 and 2010 Mexican Census of Population and Housing, and the 1995 and 2005 Count of Population and Housing collected by the Mexican National Institute of Statistics and Geography (in Spanish, Instituto Nacional de Estadística y Geografía, INEGI). First, I use official tabulations of the full-count 2000 and 2010 Mexican censuses, and the 1995 and 2005 counts available at INEGI’s website to calculate population growth (with gender and age breakdown) at municipality level. Second, I also use official tabulations of the full count based on questions included in the 2000, 2005, and 2010 population census and count regarding individuals’ location of residence five years prior to the survey. These data allow me to observe migration flows over 1995–2000, 2000–2005, and 2005–2010. Following Caballero, Cadena, and Kovak (2019), I define the return migration rate as the number of returning migrants to a municipality, divided by the municipality’s population in the survey year. Addi- tionally, I use data on migration intensity from the Mexican Population Council (CONAPO), which has information on the percentage of households whose member(s) have emigrated or returned to the United States during 1995–2000 and 2005–2010. Third, I measure internal migration flows. A municipality’s out-migration rate between t − 5 and t is the number of individuals leaving municipality i as a share of municipality i’s population in t − 5, while a municipality’s in-migration rate is the number of in-migrants as a share of i’s population in t − 5. The main caveat of using INEGI’s tabulations is that they only allow me to calculate in-migration and out-migration rates for each municipality based on individuals state of residence five years prior to the survey. Consequently, I do not observe migration rates between municipalities and within states when using INEGI tabulations of the full counts. Fourth, because movement across minor administrative divisions (i.e., municipalities or municipios) is only available for the long-form survey, I have information for an approximately 10 percent sample of the Mexican population for years 2000 and 2010. These data come from the Integrated Public Use Microdata Series (IPUMS) International, collected by the Minnesota Population Center. I also use IPUMS microdata to explore the heterogeneity in population response across different educational groups and sectors of employment. 2.2. Cross-Municipality Exposure to Trade To measure the initial industry employment shares, I use the 1999 Mexican Economic Census (with refer- ence period 1998). I also use data from the 2004 and 2009 Mexican Economic Censuses (with reference periods 2003 and 2008, respectively) to examine changes in manufacturing employment and wages over the period. Data to compute the tariffs gaps (described in detail in the next section) come from Feenstra, Romalis, and Schott (2002). Data on international trade flows are from the United Nations Comtrade Database. These data are matched to 4-digit time-consistent manufacturing industries in the Mexican Economic Census using the concordance in Pierce and Schott (2009, 2016) between the UN Comtrade 6-digit Harmonized System (HS) and the 4-digit North American Industry Classification System (NAICS, or SCIAN in Spanish). I use the data set provided by the authors to create a 4-digit industry time-invariant family-level data set containing 83 constant manufacturing industries. Finally, I use two additional sources of data to examine the main channels through which trade expo- sure affected manufacturing employment in export-oriented locations: first the Maquila Export Industry Statistics (EIME) which concluded in 2006, and second the Manufacturing Industry, Maquila and Export Services (IMMEX) which collects statistics on the export industry since 2007. These state-level statistics The World Bank Economic Review 309 from the maquiladora industry are publicly available for 17 (of 32) states that are covered by plant-level surveys on maquiladoras conducted by INEGI. 3. Empirical Strategy In this section I discuss how I construct measures of Mexican municipalities’ exposure to the changes in Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 China–US trade policy. I detail the specifications I use to estimate the causal effect of increased interna- tional competition on local employment, population growth, and migration. 3.1. Labor-Market Shock in Mexico—PNTR My primary empirical approach exploits a change in trade policy between the United States and China that generated plausible exogenous variation in Mexican export demand from the United States. The Mexican manufacturing sector experienced a rapid export-led expansion between the years 1986 and 2000, which started with the country’s entry into the General Agreement on Tariffs and Trade (GATT) in 1986 and culminated with the signing of the North American Free Trade Agreement (NAFTA) in 1994 and its implementation. The export-to-GDP ratio rose from 14 percent in 1986 to 25 percent in 2000, as Mexico became integrated into the world economy. Manufacturing exports represented 10 percent of merchandise exports over the 1980s, 43.5 percent in 1990, and 85 percent in 2000. Within NAFTA, Mexico had developed a comparative advantage in the production of labor-intensive goods (Feenstra and Kee 2007; Gallagher, Moreno-Brid, and Porzecanski 2008; Hanson and Robertson 2008; Chiquiar, Covarrubias, and Salcedo 2017). In October 2000, the United States granted China Permanent Normal Trade Relations (PNTR), which reduced uncertainty regarding potential tariff rates on Chinese exports to the United States. Before China’s accession to the WTO in 2001, the provision of tariffs rates was subject to annual renewal by the US Congress. Hence, Chinese firms faced considerable uncertainty regarding future costs of exporting. Fol- lowing China’s accession to the WTO, the US Congress voted to grant NTR rates on a permanent basis. Pierce and Schott (2016) measure the impact of PNTR as the rise in US tariffs on Chinese goods that would have occurred in the event of a failed annual renewal of China’s NTR status (i.e., non-NTR tar- iffs). They define this difference between the observed NTR tariff rates and the potential non-NTR rates in industry j as the “NTR gap”: NTRGap j = NonNTRRate j − NTRRate j . I use Pierce and Schott (2016)’s approach to construct a measure of Mexican industries’ exposure to China receiving PNTR.6 The intuition behind using this measure is that Mexican municipalities with in- dustries that benefited from NAFTA, developing a comparative advantage and increasing exports to the United States, were more negatively affected by the trade liberalization between China and the United States. The change in trade policy between China and the United States was not correlated with Mex- ican pre-existing outcomes at the local level, while the industry–municipality shares predict changes in employment through the changes in the trade policy between third countries. China’s sharp increase in exports to the United States following PNTR is particularly relevant for Mex- ican manufacturing firms, given that nearly half of the manufacturing exports are produced by maquilado- ras, or export assembly plants, with the United States as their export main destination (Utar and Torres Ruiz 2013). To illustrate China’s penetration in the US market, fig. 1 shows that China surpasses Mexico’s share of US imports shortly after the passage of the bill that granted PNTR to China in October 2000. 6 Following Pierce and Schott (2016, 2020), NTR gaps are defined only for industries whose output is subject to import tariffs in the manufacturing sector. Industries whose output is not subject to tariffs, such as service industries, are assigned NTR gaps of zero. 310 Fernández Guerrico Figure 1. Share of China and Mexico in United States’ Imports Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Source: Author’s analysis based on data from the United Nations Comtrade Database. Note: This figure shows that China surpasses Mexico’s share of US imports shortly after the United States granted PNTR to China. To provide direct evidence of the effect of PNTR as a source of increased Chinese trade competition for Mexican manufacturing exports in the US market, I estimate equation (1) separately for the two US trade partners China and Mexico: Origin Imports jt = β0 + β1 NTRGap j ∗ Postt + ν j + ηt + jt , (1) ImportsAll jt where the outcome of interest is the Chinese or Mexican import share in total US manufacturing imports in industry j and year t. The variable Postt = 1[t > 2000] is a dummy that equals 1 after 2000, and ν j and ηt are industry and year fixed effects. Table 1 shows the results both for 4-digit industries and 6-digit industries. As expected, β 1 has a different sign for each country. The negative sign of the coefficients in Columns 1 and 2 imply that PNTR is associated with decreased Mexican import penetration in the United States, while the positive sign in the coefficients in Columns 3 and 4 imply that PNTR leads to increased Chinese import penetration in the United States. Given that my outcomes of interest are at municipality level, I construct a geographically based measure of international competition. I create a municipality-level measure of the NTR gap following Pierce and Schott (2020), who compute US county-level exposure to PNTR. I construct a measure of Mexican mu- nicipalities’ (indirect) exposure to PNTR as the employment-share-weighted-average of NTR gaps across manufacturing industries that are subject to tariffs: L ji NTRGapi = NTRGap j , (2) Li j where Lji represents the employment in industry j in Mexican municipality i and Li represents total em- ployment in municipality i. Data to compute NTR gaps for each industry j using ad valorem equiva- lent tariff rates are provided by Feenstra, Romalis, and Schott (2002). I follow Pierce and Schott (2016) and use NTR gaps in 1999, immediately preceding the policy change. Industry-level employment by The World Bank Economic Review 311 Table 1. PNTR and US Import Shares—1996–2010 (1) (2) (3) (4) USImportsMEX USImportsCHINA jt jt USImportsAll USImportsAll jt jt NTRGapj *Postt −0.117*** −0.0968*** 0.425*** 0.284*** Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (0.0165) (0.00777) (0.0440) (0.0123) Rescaled 25th–75th pctile −0.0405*** −0.0365*** 0.1475*** 0.1071*** (0.0057) (0.0029) (0.0153) (0.0046) Observations 1,245 62,036 1,245 62,036 Industry j 4-digit 6-digit 4-digit 6-digit Source: Author’s analysis based on data from the United Nations Comtrade Database, the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of the relationship between China receiving PNTR and US import shares from Mexico (Columns 1 and 2) and China (Columns 3 and 4). The dependent variable is the Chinese or Mexican import share in total US manufacturing imports in industry j (4-digit level in Columns 1 and 3; 6-digit level in Columns 2 and 4) and year t (1996–2010). The variable NTRGapj is the difference between the observed NTR tariff rates and the potential non-NTR rates in industry j. The third-to-last row presents rescaled estimates to reflect the change in import shares for an industry j at the 75th compared to the 25th percentile of exposure to the tariff gap. Robust standard errors are in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). Figure 2. Cross-Municipality Exposure to the PNTR—Mexico Source: Author’s analysis based on data from the United Nations Comtrade Database, the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This map of Mexico shows the cross-municipality exposure to Chinese competition in the US market given by the employment-share-weighted-average of Normal Trade Relations (NTR) gaps across manufacturing industries that are subject to tariffs. municipality is from the 1999 Mexican Economic Census. There are 2,382 municipalities in my data, spanning the entire country. Figure 2 shows a map of Mexico’s cross-municipality exposure to PNTR. Across municipalities, the unweighted NTR gap averages 7.8 percent and has a standard deviation of 6.5 percent, with an interquartile range from 2.8 to 10.6 percent. The average employment-share-weighted 312 Fernández Guerrico NTR gap is 0.2676 with a standard deviation of 0.099, and an interquartile range from 0.2082 to 0.3246.7 Figure S4.1 in the supplementary online appendix shows the employment-share-weighted-average NTR gaps across 4-digit NAICS industries in Mexico. I exploit cross-municipality variation in exposure to PNTR based on their initial industry specializa- tion, by comparing municipalities facing high and low Chinese competition in the United States before and after China was granted PNTR. Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 3.2. Estimation In this section I explain the specifications I use to estimate the effect of exposure to international compe- tition on municipality-level population adjustment, and I discuss the required specification assumptions. Municipalities more and less exposed to international competition differ in level and trend before the change in trade policy, meaning that any direct comparison of exposed and non-exposed municipalities could be biased. To address pre-existing differences and to be able to explore the within-municipality variation in population growth, migration rates, and labor-market outcomes, I start by assuming that Yi,t = αi + δt + β Zi ∗ Postt + t · Xi γ + ui,t , (3) where Yi,t is the outcome for municipality i and year t related to population adjustment, Zi is a measure of labor-market changes at municipality level, and Postt = 1[t > 2000] is a dummy variable that equals 1 after 2000. The variables α i and δ t are unobserved municipality and time effects respectively, t · Xi is a trend for municipality i, and ui,t is the error term. I estimate quinquennial-specific models equivalent to fixed effects regressions because my data is avail- able at five-year intervals (e.g., I estimate the effect of PNTR on population and migration over two periods: 2000–2005 and 2005–2010). Taking first differences of equation (3), I obtain the regression model I will use throughout the analysis to estimate how exposure to Chinese competition based on municipalities’ initial industrial specialization affected mobility in Mexico: Yi,t = β0 + β1 Zi + Xi γ + ui,t , (4) where Yi,t represents the change in the outcome variable in municipality i between years t − 5 and t. Equation (4) will consistently estimate the casual effect of Zi , under the assumption that municipal- ities more and less exposed to the change in China–US trade policy would have had common changes in outcomes in the absence of the trade shock. Because the model is estimated in first differences, the quinquennial-specific models are equivalent to fixed effects regressions.8 I use NTRGapi , defined in equation (2), as a plausible exogenous measure of labor-market changes at municipality level, represented by Zi in equation (4). Mexican municipalities with a larger initial share of employment in industries where Chinese exports to the United States increased as a consequence of PNTR, have higher exposure to international competition. Appendix S1 shows that my results are robust to using an alternative measure of exposure to inter- national competition following the empirical approach proposed in Autor, Dorn, and Hanson (2013) instead of PNTR. Appendix S2 shows that my results are robust to using commuting zones (CZ) as the geographic unit in analysis instead of municipalities. Appendix S3 shows that my results are robust to controlling for differential violence trends at municipality level. 7 In other words, the average worker in municipalities at the 75th percentile of exposure worked in an industry with an NTR gap that was 11.64 percentage points higher than the average worker in municipalities at the 25th percentile. In the municipality-level analysis, I multiply the regression coefficients by 0.1164, the magnitude of an interquartile shift in a municipality exposure to PNTR. 8 Estimating equation (4) as a fixed-effects regression assumes that the errors are serially uncorrelated (Autor, Dorn, and Hanson 2013). The variable Xi allows for the possibility that the relationship between the outcome variable and municipality i’s baseline characteristics changes in the post-PNTR period (Pierce and Schott 2016). The World Bank Economic Review 313 Table 2. International Competition and Five-Year Changes in Log Working-Age Population (1) (2) (3) All Men Women Panel A: log (population) 2000–2005 NTRGapi 0.106*** 0.133*** 0.0853** Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (0.0375) (0.0411) (0.0364) Moving municipality from 25th to 75th pctile 0.0123*** 0.0155*** 0.0099** (0.0044) (0.0048) (0.0042) Panel B: log (population) 2005–2010 NTRGapi −0.143*** −0.179*** −0.106*** (0.0434) (0.0497) (0.0397) Moving municipality from 25th to 75th pctile −0.0166*** −0.0208*** −0.0123*** (0.0051) (0.0058) (0.0046) Observations 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table shows the effect of Mexican municipalities’ exposure to international competition on population growth over 2000–2005 (Panel A) and 2005–2010 (Panel B). The dependent variable is the change in log municipality working-age population. The variable NTRGapi is a measure of Mexican municipalities’ exposure to the change in trade policy between the United States and China. Column 1 shows changes in log total population, while Columns 2 and 3 present the results for men and women, respectively. The second-to-last row in each panel presents rescaled estimates to reflect the change in log population for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). 4. The Effect of PNTR on Aggregate Population Growth To explore aggregate changes in population growth, I use quinquennial (half-decadal) municipality-level population data from the Mexican Population Census and Population Counts. I estimate equation (4), where the outcome variable Yi,t represents the change in municipality i’s log population between years t − 5 and t. I use the NTR gap, defined in equation (2), as a plausible exogenous measure of labor- market changes at municipality level, represented by Zi in equation (4). This specification differences out any time-invariant municipality treats, and controls for pre-trends in population growth. I include the lagged five-year change in log population between 1995 and 2000—predating the change in trade policy—to control for the possibility that more and less exposed municipalities experienced differential population growth on average throughout this period (Bartik 2018; Greenland, Lopresti, and McHenry 2019; Caballero, Cadena, and Kovak 2021). Table 2 presents estimation results of the average change in municipalities’ log working-age population over 2000–2005 and 2005–2010. In addition to total population counts in Column 1, I present results by gender in Columns 2 and 3. Panel A shows the change in log working-age population between 2000 and 2005. An interquartile increase in the NTR gap increased population growth by 0.0123 log points or a 1.2 percent increase. However, Panel B shows that moving a municipality from the 25th to the 75th percentile of exposure is estimated to decrease population growth by −1.7 percent over 2005–2010. The estimates are statistically significant at the 1 percent level. The estimates presented in table 2 imply an increase in log working-age population growth in the short term (i.e., 2000–2005), followed by a decrease in population growth in the middle term (i.e., 2005–2010) among municipalities with a higher average NTR gap (i.e., those more exposed to Chinese competition in the US market). These results are in line with those documented by Greenland, Lopresti, and McHenry (2019), who find that the majority of the negative population response to PNTR in the United States occurs at a lag of seven years or more after the policy shift. The authors report that an interquartile increase in Chinese import competition exposure in the United States reduced local working-age population by 314 Fernández Guerrico −0.015 log points. This result is also consistent with previous work finding that full reallocation takes up to a decade or more (Artuç, Chaudhuri, and McLaren 2010; Dix-Carneiro 2014). However, in the short term the population adjustment in Mexico differs from that in the United States: while Greenland, Lopresti, and McHenry (2019) find a negative but relatively muted response to PNTR in the short term, I document an initially increased population growth in more exposed municipalities. All in all, these results suggest that the timing of adjustment is important for estimation and that the mechanisms behind the Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 population response to a trade-induced labor-demand shock might differ in a middle-income country like Mexico compared to the United States.9 There are several factors potentially explaining the sluggish dynamic response to the labor-demand shock that I find. First, the negative economic shock may cause declines in local income that reduce migration in the short term. Second, workers with different skill sets might be more or less mobile. Third, transitions across employers and sectors could be a mechanism by which workers adjust as opposed to regional mobility. Fourth, there could be increased return migration from the United States, where the same industry-specific shock took place concurrently. In the next section, I empirically explore migration flows with the aim of providing some more insight into these channels, and I discuss how my results can be placed within the large structural literature that estimates models of location choice.10 5. The Effect of Exposure to PNTR on Migration I observe migration in 1995–2000 and 2005–2010 based on responses to the 2000 and 2010 Mexican Population Census, and in 2000–2005 based on (a subset of) responses to the 2005 Population Count.11 I estimate equation (4), where the outcome variable Yi,t represents the in-migration, out-migration, or return migration rate in municipality i in 2000–2005 and 2005–2010. The variable NTRGapi , defined in equation (2), is the exogenous measure of labor-market changes at municipality level, represented by Zi in equation (4). I include pre-shock migration rates from municipality i between 1995 and 2000 to account for the possibility that more and less exposed municipalities experienced differential migration on average throughout this period. This specification consistently estimates the effect of exposure to international competition if exposure to PNTR is independent of potential outcomes conditional on the lagged five- year migration rate. 5.1. Return Migration from the United States In table 3 I present estimates of the effect of exposure to PNTR on return migration rates. Return migrants are defined as individuals living in Mexico during the year t, when the survey took place, but who lived in another country five years before. The return migration rate is the number of migrants divided by the source’s population in the year t − 5. Column 1 shows that moving from the 25th to the 75th percentile of municipality exposure is estimated to decrease the return migration rate by −0.02 percentage points 9 The 5 to 10 year lag after PNTR with China corresponds with the spike in homicides that took place in Mexico over 2007–2010. In appendix S3, I show that my results are robust to controlling for differential violence trends at munici- pality level. 10 In canonical spatial equilibrium models (Rosen 1979; Roback 1982), workers will migrate across locations (and sectors) until real wages are equalized, and negative labor shocks will not cause heterogeneous effects on workers based on their original locations, sectors, or occupation. Topel (1986) and Moretti (2011) highlight that moving costs alter this pre- diction, leading to heterogeneous incidence of local labor-demand shocks on directly exposed workers. Moreover, there is a large literature that estimates models of local or sector/occupation choice allowing for moving costs across loca- tions (Kennan and Walker 2011; Bishop 2012; Diamond 2016; Morten and Olivera 2016; Shenoy 2016), across sectors (Artuç, Chaudhuri, and McLaren 2010; Dix-Carneiro 2014), and occupations (Artuç and McLaren 2015; Traiberman 2019). See Bartik (2018) for a discussion. 11 Migration data over the period 2000–2005 are not available to calculate all rates. The World Bank Economic Review 315 Table 3. International Competition and Five-Year Changes in Return Migration Rates (1) (2) (3) All Men Women Panel A: Return migration rate from United States 2000–2005 NTRGapi −0.00188*** −0.00324*** −0.000639 Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (0.000676) (0.00101) (0.000434) Moving municipality from 25th to 75th pctile −0.0002*** −0.0004*** −0.0001 (0.0001) (0.0001) (0.0001) Panel B: Return migration rate from United States 2005–2010 NTRGapi −0.0222*** −0.0337*** −0.0116*** (0.00323) (0.00517) (0.00168) Moving municipality from 25th to 75th pctile −0.0026*** −0.0039*** −0.0014*** (0.0004) (0.0006) (0.0002) Observations 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of the relationship between China receiving PNTR, which increased Mexican municipalities’ exposure to Chinese competition in the US market, and return migration rates over 2000–2005 (Panel A) and 2005–2010 (Panel B). The dependent variable is the return migration rate from the United States. The variable NTRGapi is a measure of Mexican municipalities’ (indirect) exposure to the change in trade policy between the United States and China. The second-to-last row presents rescaled estimates to reflect the change in the return migration rate for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). Table 4. International Competition and Change in Percentage of Households with Migrants 2005–2010 (1) (2) % Households with % Households with migrants to United States returned migrants NTRGapi 2.672** −3.020*** (1.317) (0.713) Moving a municipality from 25th to 75th pctile 0.3111** −0.3515*** (0.1533) (0.0830) Observations 2,382 2,382 Source: Author’s analysis based on data from the Mexican Population Council, the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table shows the effect of Mexican municipalities’ exposure to international competition in the US market on Mexico–US migration over 2005–2010 with respect to the pre-shock period 1995–2000. The dependent variable is the percentage of households with migrants to the United States in Column 1 and the percentage of households with returned migrants from the United States in Column 2. The variable NTRGapi is a measure of Mexican municipalities’ exposure to the change in trade policy between the United States and China. The second-to-last row presents rescaled estimates to reflect the change in returned migration for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). for the overall population between 2000 and 2005 and −0.3 percentage points between 2005 and 2010. Columns 2 and 3, which present the results for men and women, respectively, imply that the decrease in return migration is driven by men. The estimates in Column 3 of female return migration rates are imprecisely estimated in the short term, and they are relatively smaller in magnitude in the medium term. To further explore the migration response to and from the United States, I use additional data on US–Mexico migration from the Mexican National Population Council (CONAPO). This data set has information on the percentage of households with migrants to the United States and the percentage of households with return migrants in 1995–2000 and 2005–2010. Table 4 shows that moving a munic- ipality from the 25th to the 75th percentile of exposure to trade competition increased the percentage 316 Fernández Guerrico of households with migrants to the United States by 0.31 percentage points over 2005–2010, whereas it decreased the percentage of households with return migrants by −0.35 percentage points.12 All in all, these estimates suggest that return migration is not the driver of the increased population growth observed in the first five years after the United States granted PNTR to China. Thus, if Mexican workers returned to Mexico as a consequence of the negative effect of PNTR in US local labor markets (where the same industry-specific shock took place concurrently), they were less likely to choose more Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 exposed municipalities in Mexico. Turning to the effects over the medium term, if Mexican workers re- turned to Mexico, in response for example to the Great Recession, they were less likely to choose Mexican municipalities exposed to PNTR.13 The decline in return migration 6 to 10 years after the negative labor- demand shock may have contributed to the decreased population growth in the medium term, though in the next section I show that internal migration played a larger role. 5.2. Internal Migration in Mexico Workers whose initial municipalities faced greater competition may choose to migrate out to less affected regions. Similarly, more exposed municipalities might be less likely to attract migrants after a negative labor-demand shock. Previous work finds mixed migration responses to trade shocks at the local level. In the United States, Autor, Dorn, and Hanson (2013) find that rising imports per worker due to China’s emergence were not clearly associated with population growth at the local level. Greenland, Lopresti, and McHenry (2019) confront these results by documenting that an interquartile increase in Chinese import competition exposure in United States reduced local working-age population by 0.015 log points, with the majority of the population response taking place in the medium term (i.e., 7 years or more). While results in Greenland, Lopresti, and McHenry (2019) suggest that young people tended to move away (i.e., out- migrate) from trade-shocked locations, Monras (2018) documents that most of the response of internal migration in the United States during the Great Recession is accounted for by variation in in-migration. The previous evidence is also mixed in less developed economies. For example, Majlesi and Narciso (2018) find that a 1 standard deviation increase in exposure to competition from China is associated with a 1 percentage point higher probability of out-migration from 150 municipalities covered in the Mexican Family Life Survey.14 However, Dix-Carneiro and Kovak (2017) and Dix-Carneiro and Kovak (2019) find no evidence for systematic migration responses to liberalization-induced labor-demand shocks in Brazil during the 1990s. In tables 5 and 6, I present estimation results of the effect of Mexican municipalities’ exposure to increased Chinese competition in the US market on in-migration and out-migration, respectively. Table 5 estimates equation (4) using the in-migration rates over the 2000–2005 period and over the 2005–2010 period as dependent variables in Panels A and B, respectively. I include the in-migration rate during the 1995–2000 period to control for pre-shock migration rates (Bartik 2018; Caballero, Cadena, and Kovak 2019; Greenland, Lopresti, and McHenry 2019). Column 1 shows that moving from the 25th to the 75th percentile of municipality exposure is estimated to decrease the overall in-migration rates by −0.05 percentage points over 2000–2005, although this change is imprecisely estimated, and −0.2 percentage 12 Given that there are no data on migration over 2000–2005, the results presented in table 4 are comparable to those in Panel B of table 3. 13 Caballero, Cadena, and Kovak (2021) find that the decline in US employment brought about by the Great Recession is associated with increase return migration and decreased emigration in Mexican locations with strong initial ties to the hardest hit US migrant destinations. These municipalities experienced 2.1 percentage points faster population growth over 2005–2010. Caballero, Cadena, and Kovak (2018) show that the average Mexican state’s return migration rate nearly quadrupled over this time period (rates were 0.3 percent on average in 2005 and 1.13 percent on average in 2010. 14 Alternatively, a change from zero to full exposure to competition from China raises the probability of migration to another municipality by 20 percentage points; their estimates imply that exposure to trade with China can explain around 10 percent of migration within Mexico between 2002 and 2005 (Majlesi and Narciso 2018). The World Bank Economic Review 317 Table 5. International Competition and Five-Year Changes in In-Migration Rates (1) (2) (3) All Men Women Panel A: In-migration rate from other states 2000–2005 NTRGapi −0.00419 −0.00753 −0.000823 Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (0.00706) (0.00733) (0.00693) Moving a municipality from 25th to 75th pctile −0.0005 −0.0009 −0.0001 (0.0008) (0.0009) (0.0008) Panel B: In-migration rate from other states 2005–2010 NTRGapi −0.0246*** −0.0298*** −0.0195** (0.00928) (0.00949) (0.00922) Moving a municipality from 25th to 75th pctile −0.0029*** −0.0035*** −0.0023** (0.0011) (0.0011) (0.0011) Observations 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of the relationship between China receiving PNTR, which increased Mexican municipalities’ exposure to Chinese competition in the US market, and in-migration rates over 2000–2005 (Panel A) and 2005–2010 (Panel B). The dependent variable is the in-migration rate to municipality i from municipalities in a different state. The variable NTRGapi is a measure of Mexican municipalities’ (indirect) exposure to the change in trade policy between the United States and China. The second-to-last row presents rescaled estimates to reflect the change in the in-migration rate for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). Table 6. International Competition and Five-Year Changes in Out-Migration Rates (1) (2) (3) All Men Women Panel A: Out-migration rate to other states 2000–2005 NTRGapi −0.0626** −0.0809*** −0.0450* (0.0253) (0.0254) (0.0257) Moving a municipality from 25th to 75th pctile −0.0073** −0.0094*** −0.0052* (0.0030) (0.0030) (0.0030) Panel B: Out-migration rate to other states 2005–2010 NTRGapi 0.0979*** 0.102*** 0.0943*** (0.0256) (0.0266) (0.0249) Moving a municipality from 25th to 75th pctile 0.0114*** 0.0118*** 0.0110*** (0.0030) (0.0031) (0.0029) Observations 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of the relationship between China receiving PNTR, which increased Mexican municipalities’ exposure to Chinese competition in the US market, and out-migration rates over 2000–2005 (Panel A) and 2005–2010 (Panel B). The dependent variable is the out-migration from municipality i to municipalities in a different state. The variable NTRGapi is a measure of Mexican municipalities’ (indirect) exposure to the change in trade policy between the United States and China. The second-to-last row presents rescaled estimates to reflect the change in the out-migration rate for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). points over 2005–2010. The latter estimate implies a −7 percent reduction with respect to the baseline in-migration rate. Similarly, table 6 presents out-migration rates over the same periods. Column 1 in Panel A shows that moving from the 25th to the 75th percentile of municipality exposure to trade is estimated to decrease 318 Fernández Guerrico overall out-migration rates by −0.7 percentage points over 2000–2005, which represents a −6 percent reduction with respect to the baseline out-migration rate. Conversely, Column 1 in Panel B shows an increase in the overall out-migration rate of 1.1 percentage points over 2005–2010, which represents a 9 percent increase with respect to baseline. Columns 2 and 3 in tables 5 and 6 present the estimates for men and women, respectively. Results by gender are similar in magnitude over the period 2005–2010 (Panel B). However, changes in migration rates are larger in magnitude for men over the period 2000–2005 Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (Panel A). The results presented thus far provide some insights into how migration responses explain the popu- lation growth in municipalities more exposed to PNTR. In the five years following the change in trade policy, more exposed municipalities increased population growth, driven by declines in out-migration. I find evidence against these results being driven by return migration from the United States, which also rela- tively declines in more exposed areas. In-migration is negative, although imprecisely estimated in the short term. Conversely, 6 to 10 years after the plausibly exogenous change in trade policy, exposure to increased trade competition is associated with decreased population growth, driven by declines in in-migration and return migration rates, and increased out-migration. My results are consistent with lagged population adjustments that are driven by significant changes in internal migration, documented in the related literature.15 For example, Bartik (2018) finds that ex- posure to PNTR is associated with slightly higher out-migration rates and lower in-migration rates over 2000–2010 in the United States. Moving from the 25th to 75th percentile of CZ exposure is estimated to increase out-migration rates by 0.1–0.4 percentage points, although this change is imprecisely estimated, and decrease in-migration by −0.8 percentage points.16 The paper highlights workers’ imperfect mobility to move locations in response to either positive (hydraulic fracturing) and negative (PNTR with China) changes in labor demand, which suggests that migration responses are not necessarily driven by losses of industry-specific human capital or other features of negative shocks in particular. However, liquidity- constrained individuals who migrate to cope with a negative shock might be less likely to invest in mi- gration compared to those who accumulate positive shocks and save up for migration over time, as an investment to benefit from higher wages in a further-away city (Kleemans 2015). If moving costs prevent workers from relocating in response to a negative labor-demand shock, the migration response may be heterogeneous across sub-populations. In the next section I examine whether less educated workers and manufacturing workers drive the migration response documented thus far. 5.3. Heterogeneous Effects of Exposure to PNTR Structural models of location choice examine two different types of explanations for the incidence of local labor-demand shocks—one based on mobility costs and one based on compensating factors. If out- migration of workers is low primarily because of mobility costs, then the incidence of local labor-demand shocks will be primarily borne by low-skilled workers, who are comparatively immobile. Alternatively, 15 A large literature investigates the effects of local labor-demand shocks and worker migration responses. In the United States, this literature has documented significant population responses to changes in local wages (Topel 1986; Blanchard and Katz 1992; Moretti 2011; Bartik 2018). 16 In addition to studying the negative effect of PNTR exposure, Bartik (2018) studies the effect of a positive local income shock due to the fracking boom in the United States over the same period. The author finds that fracking reduces out- migration by −1.5 percentage points for non-college individuals. This represents roughly a −7 percent decline relative to the baseline out-migration rate of 19 percent. The estimate for in-migration is positive and of moderate magnitude, although imprecisely estimated. Despite the large rise in local income, fracking has not caused large long-term increases in in-migration, although the estimated decline in out-migration is moderate in magnitude. This muted in-migration result is consistent with the finding that fracking has large effects on the earnings of the original residents of exposed locations. The World Bank Economic Review 319 Table 7. International Competition and Five-Year In-Migration Rates (IPUMS Sample) (1) (2) (3) (4) (5) (6) (7) All Men Women Less than Completed Manufacturing Other high school high school sector sectors Panel A: In-migration rate from other municipalities 2005–2010 Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 NTRGapi −0.0207 −0.0268 −0.0153 −0.0368* 0.0421 −0.102*** −0.0104 (0.0226) (0.0234) (0.0224) (0.0212) (0.0283) (0.0365) (0.0205) Moving a municipality −0.0024 −0.0031 −0.0018 −0.0043* 0.0050 −0.0119*** −0.0012 from 25th to 75th pctile (0.0026) (0.0027) (0.0026) (0.0025) (0.0033) (0.0043) (0.0024) Panel B: In-migration rate from other states 2005–2010 NTRGapi −0.0433** −0.0433** −0.0436** −0.0547*** −0.0105 −0.103*** −0.0352** (0.0183) (0.0187) (0.0181) (0.0186) (0.0184) (0.0320) (0.0157) Moving a municipality −0.0051** −0.0051** −0.0051** −0.0064*** −0.0012 −0.0120*** −0.0041** from 25th to 75th pctile (0.0021) (0.0022) (0.0021) (0.0022) (0.0022) (0.0037) (0.0018) Observations 2,382 2,382 2,382 2,382 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, the Integrated Public Use Microdata Series (IPUMS) International, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of the relationship between China receiving PNTR, which increased Mexican municipalities’ exposure to Chinese competition in the US market, and in-migration rates over 2005–2010. In Panel A, the dependent variable is the in-migration rate to municipality i from any other municipality in Mexico. In Panel B, the dependent variable is the in-migration rate to municipality i from municipalities in a different state. The variable NTRGapi is a measure of Mexican municipalities’ (indirect) exposure to the change in trade policy between the United States and China. The second-to-last row presents rescaled estimates to reflect the change in the in-migration rate for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). low-skilled workers might be less likely to out-migrate if the incidence local labor-demand shocks are mitigated by housing and social insurance programs (Notowidigdo 2020).17 To explore whether the migration response to PNTR is driven by certain sub-populations, I use a 10 percent subsample of the Mexican population who answered the long-form census surveys (i.e., IPUMS microdata sample). In years 2000 and 2010 these surveys also asked individuals whether they lived in a different municipality within the same state five years before. Table 7 presents the estimates of equation (4) for the period 2005–2010 using the IPUMS sample. As before, the regression controls for the pre-shock in- migration rate (i.e., 1995–2000).18 Panel A shows that an interquartile shift in exposure to international competition reduced overall in-migration by 0.2 percentage points over 2005–2010, although imprecisely estimated. The magnitudes of the in-migration estimates from other states (table 5, Panel B) and from other municipalities (table 7, Panel A) are very similar.19 17 Notowidigdo (2020) documents that adverse shocks reduce the costs of housing. GMM estimates of the model reveal that the comparative immobility of low-skilled workers is not due to higher mobility costs per se, but rather a lower incidence of adverse labor-demand shocks. While mobility costs constrain out-migration, they do not similarly constrain in-migration, because there are a large number of potential in-migrants with negligible mobility costs. Consequently, positive local labor-demand shocks increase population more than negative shocks reduce population. 18 As explained in the Data section, the Mexican Population Census and count only have information on aggregate internal migration flows at municipality level to/from other states. Using this data, I am able to calculate in-migration and out- migration rates for each municipality based on individuals’ state of residence five years prior to the survey. Consequently, estimation results could be a lower bound if migration rates between municipalities and within states are relevant. Information on in-migration from other municipalities is not available in the 2005 Mexican population count. The results in table 7 should be compared to Panel B (i.e., the period 2005–2010) in table 5. 19 However, the estimates of in-migration from other states using the IPUMS sample, presented in Panel B, show an increase of 0.5 percentage points, slightly higher than the aggregate results using the full sample. 320 Fernández Guerrico The IPUMS sample microdata also enable me to explore whether the migration response is driven by particular sub-populations beyond gender. In Columns 4 and 5 of table 7, I explore variation across educational groups. The dependent variable is the change in the in-migration rate of those without a high-school diploma (Column 4) and those who have completed high school (Column 5). The decline in in-migration is driven by individuals with incomplete high school. Estimates presented in Column 4 imply that an interquartile increase in the NTR gap would have reduced in-migration for those without a Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 high-school diploma between −0.4 and −0.6 percentage points. This estimate is statistically significant at 10 percent in Panel A and 1 percent in Panel B. Finally, Columns 6 and 7 show the heterogeneous response across sectors of employment. Estimates in Column 6 imply significant reductions in in-migration from manufacturing sector workers, of about −1.2 percentage points. The estimates are more than three times larger for manufacturing workers than other sectors.20 The heterogeneous response documented in Columns 4 to 7 is consistent with the fact that PNTR affected the manufacturing sector. Given that repeated cross-section data do not allow me to observe the population response among individuals displaced from their jobs due to the exposure to the trade shock, it is reassuring to observe that the response is driven by populations directly affected by it. My results are in line with those of Greenland, Lopresti, and McHenry (2019), who explore hetero- geneity across educational groups using US IPUMS microdata and find that the working-age interquartile effect of PNTR in the census data (i.e., −0.015 log point decline in working-age population) is bounded by estimated effects among those with less than a high-school diploma (−0.025 log points) and those with a high-school diploma or some college (−0.013 log points). My results are also consistent with Aldeco, Jurado, and Ramirez Alvarez (2019) who, using a different type of local shock, find a −0.27 percent decrease in inflows of less educated Mexican workers into harder-hit areas and no effect on outflows, pointing to the importance of moving costs. Relatedly, Dix-Carneiro and Kovak (2017) find that the detrimental effect of the 1990s Brazilian trade liberalization on wages grew over time, a feature that is consistent with imperfect labor mobility across regions. Among more general labor-demand shocks, this literature largely finds evidence that workers leave or avoid declining areas and move toward areas with more job opportunities (Bartik 1991; Blanchard and Katz 1992; Carrington 1996; Black, McKinnish, and Sanders 2005; Foote, Grosz, and Stevens 2019).21 Although the initial decision to explore the effects of trade exposure on aggregate changes in popu- lation growth using quinquennial municipality-level population counts is driven by data availability (to be able to match aggregate population changes to migration responses), it sheds light on the dynamics of the adjustment process. The fact that the population response to trade shocks may be sluggish and heterogeneous across sub-populations is relevant when analyzing the effects of these shocks on other so- cioeconomic outcomes. In the next section, I examine the relationship between exposure to PNTR and local labor-market outcomes to provide an insight into the channels through which increased exposure to trade affects worker mobility, such as employment opportunities and wages. I also show that the negative effect of PNTR exposure was larger in more export-oriented locations. 20 Table S4.1 shows that 79 percent of the Mexican working-age population (i.e., 20–64 years old) had not completed high school in year 2000. The table also shows that 10 percent of the Mexican working-age population were employed in the manufacturing sector in year 2000. The male-to-female ratio in the manufacturing sector was 2 in year 2000 (see table S4.2). According to the 1998 Mexican economic census, 30 percent of the labor force was employed in manufacturing. 21 See Greenland, Lopresti, and McHenry (2019) for a review. My results are also related to the literature studying how local wages, rents, and employment respond to local labor-demand shocks (Topel 1986; Bartik 1991; Blanchard and Katz 1992; Saks and Wozniak 2011; Notowidigdo 2020), and more recently to Diamond (2016), whose results suggest that endogenous local amenity changes are an important mechanism driving workers’ migration responses to local labor-demand shocks. See Allen and Arkolakis (2014), Kline and Moretti (2014), Diamond (2016), Redding and Rossi- Hansberg (2016), Caliendo et al. (2019), Howard (2020). The World Bank Economic Review 321 Table 8. International Competition and Manufacturing Employment (1) (2) (3) All Men Women Panel A: log (manufacturing) 1998–2003 NTRGapi −1.018*** −0.345* −1.755*** Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (0.191) (0.185) (0.234) Moving municipality from 25th to 75th pctile −0.1184*** −0.0402* −0.2042*** (0.0222) (0.0215) (0.0272) Panel B: log (manufacturing) 2003–2008 NTRGapi −0.968*** −0.861*** −1.319*** (0.202) (0.207) (0.220) Moving municipality from 25th to 75th pctile −0.1127*** −0.1002*** −0.1535*** (0.0235) (0.0241) (0.0256) Observations 2,382 2,382 2,382 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table shows the effect of Mexican municipalities’ exposure to international competition on manufacturing employment over 1998–2003 (Panel A) and 2003–2008 (Panel B). The dependent variable is the change in log manufacturing employment. The variable NTRGapi is a measure of Mexican municipalities’ exposure to the change in trade policy between the United States and China. Column 1 shows changes in log total manufacturing workers, while Columns 2 and 3 present the results for men and women, respectively. The second-to-last row in each panel presents rescaled estimates to reflect the change in log manufacturing employment for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). 6. The Effect of Exposure to PNTR on Local Labor Markets The primary mechanism through which trade competition might lead to aggregate population changes and internal migration is via deterioration in employment opportunities in the manufacturing sector. In this section I quantify the first-order effects of the trade shock on employment and wages. First, I show the direct effect of exposure to PNTR on overall manufacturing employment and wages using data from the Mexican Economic Census. Second, I examine whether there is a heterogeneous response to PNTR in locations with a higher concentration of export-oriented industries using data from the Mexican maquila export industry statistics from INEGI. Third, I explore whether transitions across sectors of employment are a plausible mechanism explaining the sluggish population response that I document above. 6.1. The Effect of Exposure to PNTR on Manufacturing Employment I start by documenting a deterioration in employment opportunities in the manufacturing sector in more exposed municipalities that might have lead to aggregate population changes and internal migration that I find in the previous sections. The municipality–industry level data on manufacturing employment are available in the Mexican Economic Census, while the population counts and migration data are available in the Mexican Population Census. Given that the two data sources do not perfectly overlap, I look at the relationship between exposure to PNTR and changes in manufacturing employment over 1998–2003 and 2003–2008, but the population and migration analysis is over 2000–2005 and 2005–2010.22 Table 8 shows that manufacturing employment has a negative and statistically significant relationship with municipality exposure to Chinese competition in the US market. Column 1 estimates imply that 22 While I acknowledge this limitation in the analysis, the two data sources are the best available in terms of representa- tiveness and geographic coverage, to the best of my knowledge. 322 Fernández Guerrico overall manufacturing employment in a municipality at the 75th percentile of trade exposure declined by −0.12 log points more than in a municipality at the 25th percentile over 1998–2003, and by −0.11 log points more over 2003–2008. Columns 2 and 3 show the trade-induced decline in manufacturing employment for men (−0.04 to −0.10 log points) and women (−0.15 to −0.20 log points), respectively. The economic relevance of my estimates is in line with previous work investigating the effect of Chinese competition in the US market on Mexican labor-market outcomes. Fernández Guerrico (2021) finds that Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 manufacturing employment in a municipality at the 75th percentile of trade exposure declined by −0.08 log points more than in a municipality at the 25th percentile over 1998–2003, and by −0.15 log points more over 1998–2013. Dell, Feigenberg, and Teshima (2019) also document a negative relationship be- tween increased Chinese competition in US and Mexican manufacturing employment opportunities.23 Chiquiar et al. (2017) find regionally heterogeneous effects of exposure to Chinese competition on unem- ployment in 53 Mexican metropolitan areas; increased Chinese import penetration in the United States is associated with a 1.32 percentage point increase in the unemployment rate in border areas and a 0.41 percentage point increase in non-border areas. Utar and Torres Ruiz (2013) results indicate that about half of the −1.2 percent decline in maquiladoras employment between 2001 and 2006 can be attributed to Chinese competition.24 If the competition with China in the US market is felt significantly among maquiladoras, then one expects to see the impact on maquila employment and wages. In the next section, I quantify these first- order effects to examine whether they are a mechanism explaining the population response to the change in trade policy documented in the previous sections. 6.2. The Effect of Exposure to PNTR on Export-Oriented Industries Next, I examine whether the relationship between exposure to PNTR and manufacturing employment is heterogeneous depending on how export-oriented an industry is in a particular location. My data do not allow me to distinguish between export-oriented and domestic-market-oriented plants within a munici- pality; I try to overcome this data limitation by documenting changes in manufacturing employment in locations where the Mexican maquiladora industry has a higher concentration. Maquiladoras are labor- intensive export processing plants in Mexico that are tied to the US manufacturing sector. The number of workers employed in the Mexican maquiladora manufacturing industry dropped from 1.29 million in 2000 to 1.16 million in 2005 (INEGI 2007). In terms of sales, maquiladoras’ exports to the US represented 99.7 percent of the total maquiladoras’ exports in 1993 and 94 percent in 2006.25 Previous work has doc- umented that Mexican maquiladoras and Chinese plants had very similar export baskets (Gallagher and Porzecanski 2007; Gallagher, Moreno-Brid, and Porzecanski 2008), and find that both employment and plant growth at Mexican maquiladoras were negatively affected by Chinese competition in the United States (Utar and Torres Ruiz 2013). For this part of the analysis, I use state-level statistics from the maquiladora industry that are publicly available at INEGI: first the Maquila Export Industry Statistics (EIME) which concluded in 2006, and 23 Dell, Feigenberg, and Teshima (2019) first-stage estimates imply that a ten thousand USD increase in predicted in- ternational competition per worker results in a 0.08 (1998–2013) to 0.97 (1998–2003) standard deviation decline in employment, noting that 1 standard deviation in manufacturing employment is nearly twice as large for 1998–2013 (0.23) as for 1998–2003 (0.13). 24 The magnitude of my estimates is also comparable to the seminal papers on the China shock effect on US local labor markets. Autor, Dorn, and Hanson (2013) find that Chinese import penetration explains 25 to 66 percent of the overall decline in US manufacturing employment from 2000 to 2007, or −5 to −11 percentage points of the overall −20 percent decline. Pierce and Schott (2016) find a relative decline in American manufacturing employment of −0.15 log points as a consequence of exposure to PNTR. 25 See Utar and Torres Ruiz (2013) for a detailed description of the maquiladora industry in Mexico in this period. The World Bank Economic Review 323 Table 9. International Competition and Manufacturing Employment in Maquiladoras (1) (2) (3) 2000–2005 2007–2010 2007–2012 log (maquiladora manufacturing) NTRGaps −14.22*** −1.778*** −2.502** Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 (4.653) (0.378) (0.955) Moving state from 25th to 75th pctile −0.2730*** −0.0341*** −0.0480** (0.0893) (0.0073) (0.0183) Source EIME IMMEX IMMEX Observations (states) 17 17 17 Source: Author’s analysis based on data from the Maquila Export Industry Statistics (EIME), the Manufacturing Industry, Maquila and Export Services (IMMEX), the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table shows the effect of Mexican states’ exposure to international competition on maquiladora manufacturing employment over 2000–2005, 2007–2010, and 2007–2012. The dependent variable is the change in log manufacturing employment in export processing plants surveyed by the EIME and IMMEX with a high concentration of export-oriented industries (maquiladoras). The variable NTRGaps is a measure of Mexican states’ exposure to the change in trade policy between the United States and China. Column 1 shows changes in log maquiladora workers over 2000–2005 using data from the EIME, while Columns 2 and 3 present the results for the periods 2007–2010 and 2007–2012 using data from IMMEX. The second-to-last row presents rescaled estimates to reflect the change in log manufacturing maquiladora employment for a Mexican state at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted states. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). second the Manufacturing Industry, Maquila and Export Services (IMMEX), which collects statistics on the export industry since 2007.26 Table 9, Column 1 estimates imply that an interquartile shift in state exposure to PNTR is associated with a −0.27 log point decrease in manufacturing employment in the maquila industry between years 2000 and 2005.27 This effect size is in line with the results in Utar and Torres Ruiz (2013), who document a 23 percentage point loss in manufacturing employment in maquiladora plants as a consequence of increased Chinese competition in the US market over the same period. Columns 2 and 3 show results for the 2007–2010 and 2007–2012 periods using IMMEX data. Estimates imply that moving a state from the 25th to the 75th percentile of exposure to PNTR is associated with a decrease of −0.03 and −0.05 log points, respectively. 6.3. The Effect of Exposure to PNTR on Manufacturing Wages Finally, I examine the effect on wages. I use information from the Mexican Economic Census on the annual wage bill at municipality level.28 I do not observe individual wages, which means that I cannot distinguish between changes in the average wage due to changes in wages for individual workers or to changes in the composition of workers. This data limitation may introduce a bias in the estimation. For example, if workers with lower wages are more likely to lose employment, then the observed changes in 26 Although initially restricted to the border states and the Baja California free-trade zone, since 1989 maquiladoras can be established anywhere in Mexico. The state-level information from EIME and IMMEX is available for 17 (of 32) states that are covered by these plant-level surveys conducted by INEGI: Aguascalientes, Baja California, Coahuila de Zaragoza, Chihuahua, Ciudad de Mexico, Durango, Guanajuato, Jalisco, Mexico, Nuevo Leon, Puebla, San Luis Potosi, Sonora, Tamaulipas, Yucatan. EIME and IMMEX are not compatible in the type of variables and aggregation they used. Consequently, to approximate the periods examined as closely as possible to the rest of the analysis, I show results for 2000–2005 using EIME data, and 2007–2010 and 2007–2012 using IMMEX. Also, EIME has data on the states of Zapatecas and Sinaloa but IMMEX does not; in turn, IMMEX has data on the states of Queretaro and Veracruz de Ignacio Llave but EIME does not. 27 Estimates imply losses in employment across genders between 2000 and 2005: −0.28 log points for men and −0.27 log points for women. I do not have data by gender at state level from IMMEX. 28 Unfortunately, the annual wage bill is not disaggregated by gender. 324 Fernández Guerrico Table 10. International Competition and Manufacturing-Sector Wages (1) (2) (3) (4) log (wages) log (payroll) log (wages) log (payroll) 1998–2003 2003–2008 Panel A: All municipalities Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 NTRGapi −1.423*** −1.287*** −1.021*** −1.181*** (0.273) (0.283) (0.261) (0.270) Moving a municipality from 25th to 75th pctile −0.1656*** −0.1498*** −0.1189*** −0.1374*** (0.0318) (0.0330) (0.0304) (0.0314) Observations 2,382 2,382 2,382 2,382 Panel B: Municipalities in export-intensive regions NTRGapi −1.830*** −1.755*** −0.862** −0.947** (0.382) (0.394) (0.366) (0.380) Moving a municipality from 25th to 75th pctile −0.2359*** −0.2262*** −0.1111** −0.1221** (0.0493) (0.0508) (0.0472) (0.0489) Observations 1,076 1,076 1,076 1,076 Source: Author’s analysis based on data from the Mexican National Institute of Statistics and Geography, Feenstra, Romalis, and Schott (2002), and Pierce and Schott (2009, 2016). Note: This table presents estimates of equation (4) and shows the effect of Mexican municipalities’ exposure to international competition on the municipality-level wage bill over 1998–2003 (Columns 1 and 2) and 2003–2008 (Columns 3 and 4). The dependent variable is the change in log regular wages and payroll (including social security contributions made by employers). The variable NTRGapi is a measure of Mexican municipalities’ exposure to the change in trade policy between the United States and China. Panel A shows changes in log wages and payroll in all municipalities, while Panel B presents the results for municipalities located in export-intensive states as with a large concentration of maquiladora employment. The second-to-last row in each panel presents rescaled estimates to reflect the change in log manufacturing employment for a Mexican municipality at the 75th compared to the 25th percentile of exposure to international competition. Observations are population-weighted municipalities. Robust standard errors in parentheses (*** p < 0.01, ** p < 0.05, * p < 0.1). the average wage will understate the changes in wages relative to the case in which the composition of workers is constant (Autor, Dorn, and Hanson 2013; Blyde, Busso, and Romero 2020). I estimate the effect of exposure to PNTR on two outcome variables available in the economic census data: first, the change in log wages, which only includes regular wages paid to all employees without social security contributions,29 and second the change in the log of payroll, which includes regular wages paid to workers and social security contributions paid by the employer. I use the consumer price index (INPC) from INEGI to deflate both outcomes and express wages and payroll in year 2010 constant prices. Table 10 shows estimates of the effect of exposure to PNTR on the change in log regular wages and log total payments (including regular wages and social security contributions) made by employers at municipality level. Columns 1 and 3 in Panel A show that moving a municipality from the 25th to the 75th percentile of exposure to PNTR is associated with a statistically significant decrease in municipality-level average wages of −0.16 log points over 1998–2003 and −0.12 log points over 2003–2008. The change in log payroll is similar in magnitude, −0.15 log points over 1998–2003 and −0.14 log points over 2003– 2008. In Panel B, I replicate the analysis from Panel A for the subset of export-oriented locations, that is, municipalities located in states with a high concentration of maquiladora employment that I examined above using data from the EIME and IMMEX. The effects of exposure to PNTR are larger in this subset of municipalities; an interquartile shift in exposure to PNTR in export-oriented municipalities is associated 29 A Mexican employer that fully complies with the components of the labor regulations included in the census would pay 18 percent of wages as social security contributions. Employers are required to pay social security contributions for its wage employees; in practice though, compliance with this obligation is not uniform. Workers are considered informal when their employers do not make social security contributions for them (Blyde, Busso, and Romero 2020). Related work finds that firms replace some formally hired wage employees with informally hired wage employees as a response to the China shock (Blyde, Busso, and Romero 2020; Fernández Guerrico 2021). The World Bank Economic Review 325 with a reduction in wages and payroll of about −0.23 log points over 1998–2003 and −0.11 log points over 2003–2008. The results presented in this section imply that the labor market adjusted both along the employment margin and through wage reductions in the manufacturing sector. 6.4. The Effect of Exposure to PNTR on Non-manufacturing Employment and Wages The evidence presented thus far points to several factors explaining the dynamic population response to Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 the negative labor-demand shock that I find. My results indicate that exposure to PNTR affected pop- ulation growth via a deterioration of labor-market opportunities in the manufacturing sector, which led to income loss due to job loss and lower wages. The sluggish population adjustment is consistent with the changes in internal migration that I document in the previous sections. However, the initial positive population growth remains somewhat puzzling given the consistent negative effects on manufacturing employment opportunities and income that I document during the whole period. To shed light on why there is a reversal in the population response to PNTR, I explore whether tran- sitions across sectors are a mechanism by which workers adjust as opposed to regional mobility in the short term. I find evidence indicating that Chinese competition reallocated Mexican employment from manufacturing to services in the short run, in line with Bloom et al. (2019) and Faber, Sarto, and Tabellini (2022) findings in the United States. Column 1 in table S4.3 shows that an interquartile shift in expo- sure to PNTR is associated with a 0.03 log point increase in overall non-manufacturing employment in more exposed municipalities in 1998–2003, followed by a decline of the same magnitude in 2003– 2008. Columns 2–4 show the contribution by service sub-sectors, which experience short-term increases in employment of 0.12–0.13 log points, followed by long-term declines of −0.14 log points. The service sub-sectors that explain the overall response represent 19.6 percent (wholesale, professional services, and management) and 11.9 percent (transportation, warehousing, information, finance, insurance, and real estate) of non-manufacturing jobs at baseline.30 Lastly, table S4.4 shows the effect of PNTR on the non-manufacturing-sector wage bill. In the short term, coefficient estimates for the effect on overall wages in the non-manufacturing sector are negative but imprecisely estimated. In the long term, moving a municipality from the 25th to the 75th percentile of exposure to PNTR is associated with a statistically significant decline of −0.10 log points in the overall non-manufacturing wage bill.31 All in all, there are short-term job gains in the non-manufacturing sector that partially offset the job losses in manufacturing. The fact that manufacturing employment and wages fall immediately while service-sector employment and wages are more sluggish to respond is a possible mechanism behind the reversal in out-migration and in population growth. 7. Robustness Checks My primary empirical approach follows that of Pierce and Schott (2016) who analyze the effect of the United States granting PNTR to China in October 2000 and the surge in US imports of Chinese goods that accompanied the policy change. To the best of my knowledge, this paper is the first to exploit such a change in trade policy in Mexican labor markets. However, vast previous work exploits Chinese entry to 30 The initial share of non-manufacturing employment is 70 percent. Following the same criteria as Bloom et al. (2019), I aggregate all NAICS non-manufacturing sectors into three broad categories: Column 1 shows the total contribution of non-manufacturing sectors; Column 2 shows the contribution by non-manufacturing sub-sectors 43 (wholesale), 54–56 (professional services and management); Column 3 shows the contribution by sub-sectors 48–49 (transportation and warehousing), 51 (information), and 52–53 (FIRE); Column 4 shows the contribution by other non-manufacturing sub- sectors: 11–23 (mining, utilities, construction), 46 (retail), 61–81 (education, health, entertainment, accommodation, and food). 31 The data used and the limitations described for table 10 results on manufacturing wages apply here. 326 Fernández Guerrico the WTO as a plausible source of exogenous variation in international competition in Mexico following an empirical approach inspired by Autor, Dorn, and Hanson (2013)’s seminal paper. In appendix S1, I show that my results using exposure to PNTR are similar to those following the empirical approach used in Fernández Guerrico (2021) to analyze the effect of increased trade competition on leading causes of mortality in Mexico, following Autor, Dorn, and Hanson (2013). Second, one might be concerned that municipalities in the same Mexican CZ might be part of an Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 integrated labor market in equilibrium (Caballero, Cadena, and Kovak 2021). In appendix S2, I show that results are very similar when aggregating the unit of analysis to the Mexican CZ level, showing that the choice of Mexican market aggregation does not substantially affect my findings. Third, population growth and migration may also be influenced by factors other than exposure to trade, such as violence and crime, which could confound the results in the latter period of analysis.32 In appendix S3, I control for local homicides to capture the effects of drug-related violence in Mexico during my period of analysis. Because municipality-level violence and criminal activity might have been affected by the trade-induced manufacturing employment shock (Dix-Carneiro, Soares, and Ulyssea 2018; Dell, Feigenberg, and Teshima 2019), I control differential violence trends at municipality level. I also show that my migration and population results are not driven by increased regional violence by dropping the municipalities in the most violent states over the 2005–2010 period that overlaps with the spike in homicides (i.e., 2007–2010). Finally, I show that my results are robust to dropping municipalities in border states in table S4.5. In the sensitivity checks described in this section, the economic relevance and signs of my estimates remain unchanged. In a few cases, described in detail in each section of the supplementary online appendix, the estimation is somewhat less precise (i.e., coefficients are sometimes statistically significant at 5 or 10 percent instead of 1 percent as in the main analysis). All in all, I conclude that my results are robust to a series of sensitivity tests described above and in the supplementary online appendix. 8. Conclusion This paper studies how trade-induced changes in Mexican labor demand affect population growth and migration flows at the local level. I exploit cross-municipality variation in exposure to a change in trade policy between the United States and China that eliminated potential tariff increases on Chinese imports. I show that trade competition resulting from the United States granting China Permanent Normal Trade Relations (PNTR) increased the US manufacturing import share from China and decreased the US import share from Mexico. After documenting first-order negative effects of PNTR on Mexican manufacturing exports to the US market at industry level, I construct a Mexican municipality measure of exposure to trade competition based on their industry structure. I show that Mexican municipalities specializing in industries in which China had an initial comparative advantage were more exposed to the change in trade policy. Next, I document the deterioration in manufacturing employment opportunities, which plausibly lead to income loss (due to job loss or lower wages), as the main mechanism leading to aggregate population changes and internal migration. The results presented in this paper imply dynamic population effects in response to increased municipality-level exposure to Chinese competition in the US market. In the five years following the change in trade policy, more exposed municipalities have increased population growth, driven by declines in out-migration. I find evidence against these results being driven by return migration from the United 32 Some regions in Mexico experienced a spike in the homicide rate between 2007 and 2010, which could also affect population adjustment costs (Ajzenman, Galiani, and Seira 2015; Aldeco, Jurado, and Ramirez Alvarez 2019). However, a strand of the literature finds muted effects of violence on migration in Mexico over the same period (Basu and Pearlman 2017; Utar 2021). See the appendix for a more detailed discussion of the related literature. The World Bank Economic Review 327 States, which also relatively declines in more exposed areas. The effect on in-migration is negative, al- though imprecisely estimated in this quinquennial. Conversely, 6 to 10 years after the plausibly exogenous change in trade policy, exposure to increased trade competition is associated with decreased population growth, driven by declines in in-migration and return migration rates, and increased out-migration. My results are consistent with lagged population adjustments that are driven by significant changes in internal migration. I show that transitions across sectors of employment, as opposed to short-term regional Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 adjustments, are a plausible mechanism behind the reversal in population growth and out-migration. I find that job gains in the non-manufacturing sector partially offset the manufacturing job losses in the short term. In the long term, though, exposure to PNTR is associated with declines in employment and wages across all sectors. The slack responses in service-sector employment and wages, together with high moving costs, are possible channels behind the timing of the regional adjustment. Finally, using exposure to PNTR has advantages over methods that rely on supply-driven changes in China around the time of its accession to the WTO because it is based on a specific change in trade policy. However, this approach could also present challenges because of possible trade spillovers to third coun- tries. For example, the change in US–China policy could also have affected Chinese import penetration in non-US destinations, such as Mexico, due to shared distribution channels and fixed costs of exporting. My results are robust to using supply-driven methods used in the related literature. Nevertheless, possible trade policy spillovers to third countries are a relevant consideration for future research. Data Availability Statement The information on data sources and the replication files for this article are available at https://github.com/sofia-fg/Fernandez-Guerrico_WBER2023_Replication_Files. References Aburto, J. M., H. Beltrán-Sánchez, V. M. García-Guerrero, and V. Canudas-Romo. 2016, jan. “Homicides In Mexico Reversed Life Expectancy Gains for Men and Slowed Them for Women, 2000-10.” Health Affairs (Project Hope) 35 (1): 88–95. Acemoglu, D., D. Autor, D. Dorn, G. H. Hanson, and B. Price. 2016. “Import Competition and the Great US Employ- ment Sag of the 2000s.” Journal of Labor Economics 34(S1): 141–98. Adão, R., M. Kolesár, and E. Morales. 2019. “Shift-Share Designs: Theory and Inference.” Quarterly Journal of Economics 134(4): 1949–2010. Adda, J., and Y. Fawaz. 2020. “The Health Toll of Import Competition.” Economic Journal 130(630): 1501–40. Ajzenman, N., S. Galiani, and E. Seira. 2015. “On the Distributive Costs of Drug-Related Homicides.” Journal of Law and Economics 58(4): 779–803. Aldeco, L., A. Jurado, and A. Alvarez-Ramirez. 2019. “Internal Migration and Drug Violence in Mexico.” Unpublished manuscript. Allen, T., and C. Arkolakis. 2014. “Trade and the Topography of the Spatial Economy.” Quarterly Journal of Eco- nomics 129(3): 1085–139. Angelucci, M. 2015. “Migration and Financial Constraints: Evidence from Mexico.” Review of Economics and Statis- tics 97(1): 224–28. Arthi, V., B. Beach, and W. Hanlon. 2022. “Recessions, Mortality, and Migration Bias: Evidence from the Lancashire Cotton Famine.” American Economic Journal: Applied Economics 14(2): 228–55. Artuç, E., S. Chaudhuri, and J. McLaren. 2010. “Trade Shocks and Labor Adjustment: A Structural Empirical Ap- proach.” American Economic Review 100(3): 1008–45. Artuç, E., and J. McLaren. 2015. “Trade Policy and Wage Inequality: A Structural Analysis with Occupational and Sectoral Mobility.” Journal of International Economics 97(2): 278–94. Autor, D., D. Dorn, and G. Hanson. 2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review 103(6): 2121–68. 328 Fernández Guerrico Dorn, D., and G. Hanson. 2015. “The Labor Market and the Marriage Market: How Adverse Employment Shocks Affect Marriage, Fertility, and Children’s Living Circumstances.” ———. 2019. “When Work Disappears: Manufacturing Decline and the Falling Marriage Market Value of Young Men.” American Economic Review: Insights 1(2): 161–78. Autor, D., D. Dorn, G. Hanson, and J. Song. 2014. “Trade Adjustment: Worker-Level Evidence.” Quarterly Journal of Economics 129(4): 1799–860. Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Bartik, A. W. 2018. “Worker Adjustment to Changes in Labor Demand: Evidence from Longitudinal Census Data.” Unpublished manuscript. Bartik, T. 1991. “Who Benefits from State and Local Economic Development Policies?” W. E. Upjohn Institute for Employment Research. Basu, S., and S. Pearlman. 2017. “Violence and Migration: Evidence from Mexico’s Drug War.” IZA Journal of De- velopment and Migration 7 (18). Bazzi, S., G. Hanson, S. John, B. Roberts, and J. Whitley. 2021. “Deterring Illegal Entry: Migrant Sanctions and Recidivism in Border Apprehensions.” American Economic Journal: Economic Policy 13(3): 1–27. Belot, M. V. K., and T. J. Hatton. 2012. “Immigrant Selection in the OECD.” Scandinavian Journal of Economics 114(4): 1105–28. Bishop, K. 2012. “A Dynamic Model of Location Choice and hedonic Valuation.” Mimeo, Washington University in St. Louis: 1–38. Black, D., T. McKinnish, and S. Sanders. 2005. “The Economic Impact of the Coal Boom and Bust.” Economic Journal 115(503): 449–76. Blanchard, O. J., and L. F. Katz. 1992. “Regional Evolutions.” Brookings Papers on Economic Activity 1992(1): 1–75. Bloom, N., K. Handley, A. Kurman, and P. Luck. 2019. “The Impact of Chinese Trade on U.S. Employment: The Good, the Bad, and the Debatable.” Unpublished manuscript. Blyde, J., M. Busso, V. Faggioni, and D. Romero. 2017. “The Impact of Chinese Competition on Mexican Labor Outcomes.” Blyde, J. S., M. Busso, and D. Romero. 2020. “Labor Market Adjustment to Import Competition: Long-Run Evidence from Establishment Data.” Unpublished manuscript. Borusyak, K., P. Hull, and X. Jaravel. 2022. “Quasi-Experimental Shift-Share Research Designs.” Review of Economic Studies 89 (1): 181–213. Bound, J., and H. J. Holzer. 2000. “Demand Shifts, Population Adjustments, and Labor Market Outcomes during the 1980s.” Journal of Labor Economics 18(1): 20–54. Caballero, M. E., B. C. Cadena, and B. K. Kovak. 2018. “Measuring Geographic Migration Patterns Using Matrículas Consulares.” Demography 55(3): 1119–45. ———. 2019. “The International Transmission of Local Economic Shocks through Migrant Networks.” ———. 2021. “The International Transmission of Local Economic Shocks through Migrant Networks.” NBER WP 28696. Cadena, B. C., and B. K. Kovak. 2016. “Immigrants Equilibrate Local Labor Markets: Evidence from the Great Re- cession.” American Economic Journal: Applied Economics 8(1): 257–90. 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. Callaway, B., and P. H. Sant’Anna. 2021. “Difference-in-Differences with Multiple Time Periods.” Journal of Econo- metrics 225(2): 200–30. Carrington, W. 1996. “The Alaskan Labor Market during the Pipeline Era.” Journal of Political Economy 104 (1). Chiquiar, D., E. Covarrubias, and A. Salcedo. 2017. “Labor Market Consequences of Trade Openness and Competition in Foreign Markets.” Banco de Mexico WP 2017-01. Chiquiar, D., and G. H. Hanson. 2005. “International Migration, Self-Selection, and the Distribution of Wages: Evi- dence from Mexico and the United States.” Journal of Political Economy 113 (2). Dell, M. 2015. “Trafficking Networks and the Mexican Drug War.” American Economic Review 105(6): 1738–79. Dell, M., B. Feigenberg, and K. Teshima. 2019. “The Violent Consequences of Trade-Induced Worker Displacement in Mexico.” American Economic Review: Insights 1(1): 43–58. Diamond, R. 2016. “The Determinants and Welfare Implications of US Workers’ Diverging Location Choices by Skill: 1980-2000.” American Economic Review 106(3): 479–524. The World Bank Economic Review 329 Dix-Carneiro, R. 2014. “Trade Liberalization and Labor Market Dynamics.” Econometrica 82(3): 825–85. Dix-Carneiro, R., and B. K. Kovak. 2017. “Trade Liberalization and Regional Dynamics.” American Economic Review 107(10): 2908–46. ———. 2019. “Margins of Labor Market Adjustment to Trade.” Journal of International Economics 117: 125–42. Dix-Carneiro, R., R. R. Soares, and G. Ulyssea. 2018. “Economic Shocks and Crime: Evidence from the Brazilian Trade Liberalization.” Applied Economic Journal: Applied Economics 10(4): 158–95. Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Edmonds, E. V., N. Pavcnik, and P. Topalova. 2009. “Child Labor and Schooling in a Globalizing World: Some Evi- dence from Urban India.” Journal of the European Economic Association 7(2–3): 498–507. ———. 2010. “Trade Adjustment and Human Capital Investments: Evidence from Indian Tariff Reform.” American Economic Journal: Applied Economics 2(4): 42–75. Faber, M., A. Sarto, and M. Tabellini. 2022. “Local Shocks and Internal Migration: The Disparate Effects of Robots and Chinese Imports in the US.” NBER WP 30048. Feenstra, R. C., and H. L. Kee. 2007. “Trade Liberalisation and Export Variety: A Comparison of Mexico and China.” World Development. Feenstra, R. C., J. Romalis, and P. K. Schott. 2002. “U.S. Imports, Exports and Tariff Data 1989-2001.” NBER WP 9387. Fernández Guerrico, S. 2021. “The Effects of Trade-Induced Worker Displacement on Health and Mortality in Mex- ico.” Journal of Health Economics 80. Foote, A., M. Grosz, and A. Stevens. 2019. “Locate Your Nearest Exit: Mass Layoffs and Local Labor Market Re- sponse.” ILR Review 72(1): 101–26. Gallagher, K. P., J. C. Moreno-Brid, and R. Porzecanski. 2008. “The Dynamism of Mexican Exports: Lost in (Chinese) Translation?” World Development 36(8): 1365–80. Gallagher, K. P., and R. Porzecanski. 2007. “What a Difference a Few Years Makes: China and the Competitiveness of Mexican Exports.” Oxford Development Studies 35 (2). Goldsmith-Pinkham, P., I. Sorkin, and H. Swift. 2020. “Bartik Instruments : What, When, Why, and How.” American Economic Review 110(8): 2586–624. Greenland, A., J. Lopresti, and P. McHenry. 2019. “Import Competition and Internal Migration.” Review of Eco- nomics and Statistics 101(1): 55–59. Hanson, G. H., and R. Robertson. 2008. “China and the Manufacturing Exports of Other Developing Countries.” NBER WP No 14497. Howard, G. 2020. “The Migration Accelerator: Labor Mobility, Housing, and Aggregate Demand.” American Eco- nomic Journal: Macroeconomics 12(4): 147–79. Ibarraran, P., and D. Lubotsky. 2007. “Mexican Immigration and Self-Selection: New Evidence from the 2000 Mexican Census.” In Mexican Immigration to the United States, edited by G. Borjas, Chapter 5, 159–92. University of Chicago Press. INEGI. 2007. “Estadísticas Económicas. Industria Maquiladora de Exportación.” Technical report, Instituto Nacional de Estadística, Geografía e Informática, Mexico. Kaestner, R., and O. Malamud. 2014. “Self-Selection and International Migration: New Evidence from Mexico.” Review of Economics and Statistics 96(1): 78–91. Kennan, J., and J. R. Walker. 2011. “The Effect of Expected Income on Individual Migration Decisions.” Econometrica 79(1): 211–51. Khanna, G., C. Medina, A. Nyshadham, and J. A. Tamayo. 2019. “Formal Employment and Organized Crime: Re- gression Discontinuity Evidence from Colombia.” Kleemans, M. 2015. “Migration Choice under Risk and Liquidity Constraints.” Unpublished manuscript. Kline, P., and E. Moretti. 2014. “Local Development, Agglomeration Economies, and the Big Push: 100 Years of Evidence from the Tennesse Valley Authority.” Quartely Journal of Economics: 275–331. Kovak, B. K.. 2013. “Regional Effects of Trade Reform: What Is the Correct Measure of Liberalization?” American Economic Review 103(5): 1960–76. Majlesi, K., and G. Narciso. 2018. “International Import Competition and the Decision to Migrate: Evidence from Mexico.” Journal of Development Economics 132: 75–87. Mau, K. 2017. “US Policy Spillover(?) – China’s Accession to the WTO and Rising Exports to the EU.” European Economic Review 98: 169–88. 330 Fernández Guerrico McCaig, B. 2011. “Exporting out of Poverty: Provincial Poverty in Vietnam and U.S. Market Access.” Journal of International Economics 85(1): 102–13. McKenzie, D., and H. Rapoport. 2010. “Self-Selection Patterns in Mexico-U.S. Migration: The Role of Migration Networks.” Review of Economics and Statistics 92(4): 811–21. Mendez, O. 2015. “The Effect of Chinese Import Competition on Mexican Local Labor Markets.” North American Journal of Economics and Finance 34: 364–80. Downloaded from https://academic.oup.com/wber/article/37/2/305/7058782 by Joint Bank-Fund library user on 04 September 2023 Monras, J. 2018. “Economic Shocks and Internal Migration.” Unpublished manuscript. Moretti, E. 2011. “Local Labor Markets.” Handbook of Labor Economics 4. Morten, M., and J. Olivera. 2016. “Paving the Way to Development: Costly Migration and Labor Market Integration.” NBER WP 22158. Notowidigdo, M. J. 2020. “The Incidence of Local Labor Demand Shocks.” Journal of Labor Economics 38 (3). Pavcnik, N. 2017. “The Impact of Trade on Inequality in Developing Coutnries.” NBER WP No. 23878. Pierce, J. R., and P. K. Schott. 2009. “A Concordance between Ten-Digit U.S. Harmonized System Codes and SIC/NAICS Product Classes and Industries.” NBER WP 15548. ———. 2016. “The Surprisingly Swift Decline of US Manufacturing Employment.” American Economic Review 106(7): 1632–62. ———. 2020. “Trade Liberalization and Mortality: Evidence from US Counties.” American Economic Review: In- sights 2(1): 47–64. Quiñones, E. J. 2019. “Anticipatory Migration and Local Labor Responses to Rural Climate Shocks.” Unpublished manuscript. Redding, S. J., and E. Rossi-Hansberg. 2016. “Quantitative Spatial Economics.” NBER WP 22655. Roback, J. 1982. “Wages, Rents, and the Quality of Life Wages.” Journal of Political Economy 90(6): 1257–78. Rosen, S. 1979. “Wage-Based Indexes of Urban Quality of Life.” Current Issues in Urban Economics: 1979. Saks, R. E., and A. Wozniak. 2011. “Labor Reallocation over the Business Cycle: New Evidence from Internal Migra- tion.” Journal of Labor Economics 29(4): 697–739. Shenoy, A. 2016. “Migration Decisions and Persistent Earnings Differentials: Evidence from Thailand.” Unpublished manuscript. Topalova, P. 2010. “Factor Immobility and Regional Impacts of Trade Liberalization: Evidence on Poverty from India.” American Economic Journal: Applied Economics 2(4): 1–41. Topel, R. H. 1986. “Local Labor Markets Author.” Journal of Political Economy 94 (3). Traiberman, S. 2019. “Occupations and Import Competition: Evidence from Denmark.” American Economic Review 109(12): 4260–301. Utar, H. 2018. “Workers beneath the Floodgates: Low-Wage Import Competition and Workers’ Adjustment.” Review of Economics and Statistics 100(4): 631–47. ———. 2021. “Firms and Labor in Times of Violence: Evidence from the Mexican Drug War.” SSRN Electronic Journal. Utar, H., and L. B. Torres Ruiz. 2013. “International Competition and Industrial Evolution: Evidence from the Impact of Chinese Competition on Mexican Maquiladoras.” Journal of Development Economics 105: 267–87. Wooldridge, J. M.. 2005. “Violating Ignorability of Treatment by Controlling for Too Many Factors.” Econometric Theory 21(5): 1026–28.