\documentclass[12pt]{article} \usepackage{amssymb} \usepackage{graphicx} \usepackage[dcucite]{harvard} \usepackage{amsmath} \usepackage{color} \usepackage{setspace} \usepackage{booktabs} \usepackage{hyperref} \usepackage[T1]{fontenc} \usepackage{threeparttable} \usepackage{array} \usepackage{longtable} \usepackage{pdfcomment} \usepackage{caption} \usepackage{subcaption} \usepackage{comment} \setcounter{MaxMatrixCols}{10} \newtheorem{theorem}{Theorem} \newtheorem{acknowledgement}[theorem]{Acknowledgement} \newtheorem{algorithm}[theorem]{Algorithm} \newtheorem{axiom}[theorem]{Axiom} \newtheorem{case}[theorem]{Case} \newtheorem{claim}[theorem]{Claim} \newtheorem{conclusion}[theorem]{Conclusion} \newtheorem{condition}[theorem]{Condition} \newtheorem{conjecture}[theorem]{Conjecture} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{criterion}[theorem]{Criterion} \newtheorem{definition}[theorem]{Definition} \newtheorem{example}[theorem]{Example} \newtheorem{exercise}[theorem]{Exercise} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{notation}[theorem]{Notation} \newtheorem{problem}[theorem]{Problem} \newtheorem{proposition}[theorem]{Proposition} \newtheorem{remark}[theorem]{Remark} \newtheorem{solution}[theorem]{Solution} \newtheorem{summary}[theorem]{Summary} \newenvironment{proof}[1][Proof]{\textbf{#1.} }{\ \rule{0.5em}{0.5em}} \setlength{\topmargin}{0.1in} \setlength{\headheight}{0in} \setlength{\headsep}{0in} \setlength{\topskip}{0in} \setlength{\textheight}{8.5in} \setlength{\oddsidemargin}{0in} \setlength{\evensidemargin}{0in} \setlength{\textwidth}{6.5in} \makeatletter \setlength{\@fptop}{0pt} \makeatother \begin{document} \title{Not According to Plan: Worker-level responses to the USMCA high wage labor value content rules requirement\vspace{0.5in}% } \author{Stephanie Fortune-Taylor and Ross Hallren\thanks{U.S. International Trade Commission.\newline Contact emails: Stephanie.Fortune-Taylor@usitc.gov}} \date{\vspace{1.5in}% \today} \thispagestyle{empty} { % set font to helvetica (arial) to make it 508-compliant \fontfamily{phv}\selectfont \begin{center} %{\Large Not According to Plan: \vspace{0.125in}} \\ {\Large Worker-level Responses to the USMCA High Wage Labor Value Content Rules Requirement \vspace{0.25in}} \\ \vspace{0.25in} {\large Stephanie Fortune-Taylor \\} {\large Ross J. Hallren \\} \vspace{2.0in} {\large ECONOMICS WORKING PAPER SERIES}\\ Working Paper 2022--01--A \\ \vspace{0.5in} U.S. INTERNATIONAL TRADE COMMISSION \\ 500 E Street SW \\ Washington, DC 20436 \\ \vspace{0.25in} January 2022 \end{center} \vfill \noindent Office of Economics working papers are the result of ongoing professional research of USITC Staff and are solely meant to represent the opinions and professional research of individual authors. These papers are not meant to represent in any way the views of the U.S. International Trade Commission or any of its individual Commissioners. Working papers are circulated to promote the active exchange of ideas between USITC Staff and recognized experts outside the USITC and to promote professional development of Office Staff by encouraging outside professional critique of staff research. \vspace{0.25in} \newpage \thispagestyle{empty} % remove headers, footers, and page numbers from cover page \begin{flushleft} %Not According to Plan:% Worker-level Responses to the USMCA High Wage Labor Value Content Rules Requirement\\ Stephanie Fortune-Taylor, PhD and Ross J. Hallren, PhD \\ Office of Economics Working Paper 2022--01--A\\ January 2022\\~\\ \end{flushleft} \vfill \begin{abstract} \noindent There is robust empirical evidence that overall, NAFTA has had, at most, a small aggregate effect on wages, employment, unemployment, and GDP growth. However, NAFTA’s negligible net effects mask significant negative effects on employment and wages for some industries and regions, and by gender. Under the United States–Mexico–Canada Agreement (USMCA), multiple measures were included to safeguard against U.S. job losses. One such measure is USMCA’s high wage components of the labor value content (HW-LVC) requirement. In this paper, we investigate if the announcement of these measures ameliorated the demographically asymmetric labor adjustment effects of NAFTA. Our baseline results indicate that in the anticipatory period since USMCA’s HW-LVC publishing and before the rules entered into force, overall auto industry hours worked have remained unchanged, and auto worker wages have experienced a 6.6\% increase. In contrast to these baseline estimates, we find that female production workers experienced lower wage growth than their male counterparts during the in post-publishing period before the rules entered into force. Black workers demonstrated no statistical change in wages or hours from the pre-USMCA period. The divergence of female production worker wage growth from the aggregate autoworker estimates highlight the importance of a distributional approach to analyses of worker-level welfare effects, as our aggregate worker estimates masked relevant subgroup heterogeneity. \end{abstract} \vfill \begin{flushleft} Corresponding author: Stephanie Fortune-Taylor, PhD, Office of Economics\\ \href{mailto:Stephanie.Fortune-Taylor@usitc.gov}{Stephanie.Fortune-Taylor@usitc.gov}\\ \end{flushleft} } % end of helvetica (arial) font \clearpage \newpage \doublespacing \setcounter{page}{1} \section{Introduction} There is robust empirical evidence that overall, NATFA has had, at most, a small aggregate effect on numerous U.S. welfare indicators, including wages, employment, unemployment, and GDP growth. However, NAFTA’s negligible net effects mask significant negative effects on employment and wages for some industries and regions, and by gender (\citeasnoun{Roma2007}, \citeasnoun{FrancisZheng2011}, \citeasnoun{Peterson2014}, \citeasnoun{CaliendoParro2015}, \citeasnoun{USITC2016}, and \citeasnoun{WolduETAL2018}). This geographic and industry concentration of employment displacement, especially in the automotive sector, prompted the Trump administration to renegotiate the Agreement’s terms of trade (\citeasnoun{ChatzkyETAL2020}). Under the United States–Mexico–Canada Agreement (USMCA), NAFTA’s successor, multiple measures were included to safeguard against further U.S. job losses. One such measure is USMCA’s high wage components of the labor value content (HW-LVC) requirement, which conditions vehicle producers’ preferential tariff treatment on fulfillment of an average wage benchmark. Specifically, 40-45 percent of the vehicle’s value must be produced at a facility where the hourly base rate is \$16 dollars per hour or its national equivalent (\citeasnoun{FR2020a}). The HW-LVC provision is intended to “support North American jobs”, “benefit American workers”, and “drive higher wages” in the United States by raising Mexican wages (\citeasnoun{USTR2020a}). In theory, raising Mexican auto wages to U.S. levels and thus eliminating wage savings gained by Mexican production would promote greater auto investments at home. Increases in domestic employment would follow increases in domestic investment, rejuvenating U.S. auto manufacturing. So far, however, early anecdotal evidence suggests that practice has not followed theory. On the contrary, producers have doubled-down on their commitments to do business in Mexico: some have raised hourly wages to the requisite \$16/hr., stated willingness to pay the tariff associated with missing the HW-LVC \$16 benchmark, and planned to offset the cost of tripling Mexican worker wages by installing robots to replace workers (\citeasnoun{Kim2020} and \citeasnoun{Enriquez2020}). This anecdotal divergence from the intended outcome motivates our empirical analysis. How have automotive workers fared in the anticipatory, pre-entry into force period since the announcement of the HW-LVCs? Have Blacks and females, groups who have historically experienced a wage penalty in manufacturing employment, fared differently than others? Our baseline results indicate that during the anticipatory window since USMCA’s HW-LVC publishing and before the rules' entry into force, auto worker wages experienced a 6.6\% increase. In contrast to these baseline estimates, we find that female production workers earned less than their male counterparts during the anticipatory window. Notably, female production worker wage density shifts from the pre- to post- period show that the greatest employment gains came in the annual wage range commensurate with a \$16 an hour wage, the published minimum wage under the HW-LVC. Black workers demonstrated no statistical change in wages from the pre-USMCA period.The divergence of female production worker wage growth from the aggregate autoworker estimates highlight the importance of a distributional approach to analyses of worker-level welfare effects, as aggregate worker effects masked relevant subgroup heterogeneity. In addition, as these trends began before HW-LVC implementation, estimates of the effect of the rules' implementation should incorporate these anticipatory trends into empirical analyses. This paper proceeds as follows: in section 2, we begin with a literature review covering the effect of NAFTA on the employment outcomes of American workers, both in the aggregate, and separated by auto industry, gender and “race”.\footnote{We use the term “race”, acknowledging that it is a social construct, rather than a biological distinction.} We then provide background on the high wage labor value content requirements, including the conditions motivating their inclusion in USMCA and the details governing their implementation and entry into force, in section 3. The empirical methodology follows in section 4, where we detail our identification strategy and explain our two-part econometric methodology. In sections 5 and 6, we discuss our results, propose robustness checks, and examine the economic implications of our findings. We conclude in section 7 by summarizing our findings and proposing next steps for subsequent phases of our research on the USMCA HW-LVC. \section{Literature Review \label{sec: section2}} During the NAFTA period (1994-2020), the combination of the low wage rate and increasingly comparable productivity to U.S. levels influenced industry investment and employment increases in the Mexican auto sector (Figure 1) (\citeasnoun{MarangerSwiecki2016} and \citeasnoun{KlierRubenstein2017}). During same period, U.S. auto factories closed, and U.S. auto manufacturing wages stagnated from 2002-2018 (\citeasnoun{KlierRubenstein2017} and \citeasnoun{BLS2021}). Though not diagnostic, the synchronicity of auto sector labor market growth in Mexico and market shrinkage in the U.S. raised alarm bells among manufacturing workers, politicians, and industry representatives.\footnote{The viral 2016 video of a Carrier executive informing Indiana manufacturing employees that their jobs would be offshored to Mexico became a major talking point in the 2016 election, and further reinforced the prevalent public opinion that NAFTA is in part responsible for the demise of U.S. manufacturing (\citeasnoun{Schwartz2016}).} Though we are unaware of current studies that disaggregate auto industry welfare effects by gender or race, national, sector-nonspecific effects by gender follow a similar pattern to those in the NAFTA-region auto industry: NAFTA’s establishment increased Mexican women’s wages and employment, and NAFTA’s tariff reductions raised employment and wage bill shares for Mexican females employed in blue collar jobs (\citeasnoun{Aguayo-TellezETAL2012}). For U.S. women, however, NAFTA slowed wage growth, especially for married blue collar women, and led to employment declines (\citeasnoun{SaureZoabi2014}, \citeasnoun{HakobyanMcLaren2016}, and \citeasnoun{HakobyanMcLaren2017}). The importance of our study is emphasized by the limited literature that disaggregates effects of NAFTA by race or ethnicity. \citeasnoun{Benguria2020}’s working paper estimating the effect of tariff changes under NAFTA on employment outcomes is the sole source of econometric analysis. Benguria finds that tariff liberalization under NAFTA reduced employment and increased unemployment of Nonwhite workers by more than White workers. EPI finds that, controlling for education, Black and Latino manufactures workers earn 23 and 25 percent less, respectively, than their White peers (\citeasnoun{Public2021} ). Additionally, research by Public Citizen suggests that Black workers are disproportionately affected by offshoring of transportation investment and jobs (\citeasnoun{Public2021}). This disproportionate effect on Black workers is, in part, due to the fact that during the NAFTA period (1993-2019), Black workers were over-represented in the transportation equipment industry. The United States General Accounting Office’s (GAO) Trade Adjustment Assistance (TAA) report (2000) assesses how both TAA and the North American Free Trade Agreement TAA (NAFTA-TAA) program that operated from 1994 to 2002 have “met the needs of workers affected by greater foreign trade and increased imports.” The report’s descriptive statistics for FY 1999, when compared to descriptive statistics from the 1999 the nationally representative American Civilian Labor Force (ACLF) dataset, indicate that Blacks and Hispanic/Latino workers were over-represented in the certified TAA rosters, and Hispanic/Latino workers were over-represented in the certified NAFTA-TAA roster.\footnote{Forty-seven percent of certified NATFTA-TAA participants in 1999 were Hispanic/Latino, while the Hispanic/Latino ACLF for 1999 was 10.5\%. The 1999 percentages of certified NATFA-TAA participants by race are documented in Appendix I of the GAO report. The 1999 percentages ACLF workforce participants are sourced from BLS, and are linked here.} More recently, the U.S. Department of Labor’s (USDOL) FY 2019 Annual Report of the TAA for Workers Program (2020) indicates that Blacks are over-represented among TAA program enrollees, and Whites and Hispanics are underrepresented.\footnote{Note that TAA self-reported race indicators are only available for TAA program enrollees. Of the universe of U.S. workers affected by trade, some will be covered under filed TAA petitions, and others will not. Of those covered under filed petitions, some will have their petitions certified, and some will not. Of workers covered under certified petitions, some will enroll in the TAA program and become a certified participant enrollee, and some will not. Because the sample from which TAA estimates are derived is such a filtered subset of workers potentially affected by trade, users of TAA estimates should exercise caution when generalizing the estimates to a broader context.} One of the difficulties in estimating the effect of NAFTA on the welfare of U.S. autoworkers is the need to disentangle the contributions of technological change from the contributions of trade liberalization. In the United States and Mexico, both aggregate auto industry and overall manufacturing employment fell over the NAFTA period. \citeasnoun{WenReinbold2019} note that the U.S. auto manufacturing employment decline—which trends closely with the overall U.S. manufacturing decline—is due mostly to increased worker productivity, but that reducing the vehicle trade deficit could appreciably increase auto employment. \citeasnoun{HufbauerSchott2005} estimate that the downward effect of technology on U.S. auto employment dominates the downward effect of Mexican employment. U.S. real auto compensation, which remained stagnant over the period from 1994 to 2004, is estimated to have declined due to a shift toward imports from low-wage countries. \citeasnoun{FukaoETAL2003} find that FDI inflow was the key driver influencing Mexico’s growing share of U.S. auto imports. Indeed, over the period from 1993, just before NAFTA’s entry into force, to 2016, U.S. auto imports from Mexico increased by 765\%, while exports to Mexico increased by 262\% (\citeasnoun{KoopmanETAL2010} and \citeasnoun{VillarealFergusson2017}). A combination of high U.S. production costs, easy access to inexpensive raw materials, and a Mexcan labor force of varying skill levels attracted and sustained U.S. FDI growth in the Mexican automotive sector (\citeasnoun{CuevasLopez2019}). These results validate long-held U.S. autoworker concerns of company relocations to Mexico, fueled by Mexican plants boasting “high quality workers at low wages” (\citeasnoun{CBO1993}). Indeed, even those who believe that NAFTA helped the domestic auto industry note that NAFTA resulted in auto industry “rationalization of production and hence job displacements” (\citeasnoun{BurfisherETAL2001}). USMCA’s imposition of a production price floor within the LVC provision seeks to balance the U.S.-Mexico wage inequality by giving Mexican producers a choice: tariffs or higher wages. If producers choose tariffs, U.S. workers stand to benefit. Historically, workers in tariff protected industries have benefited from tariff imposition (\citeasnoun{Gertz2020}). Trade protection increases relative wages for workers in protected industries, and when trade is liberalized, relative wages in industries facing the deepest cuts fare the worst (\citeasnoun{GoldbergPavcnik2005}). Under the Canada-U.S. Free Trade Agreement, for example, this relationship between tariff cuts and wage cuts dominated the industry-specific human capital effect (\citeasnoun{Townsend2007}). If, however, Mexican producers choose higher wages over tariffs, forfeiting the low-wage premium that make Mexico a more attractive destination than the U.S. for auto producers, U.S. workers may still benefit if producers' response to reduced wage savings is to shift production and investment to the U.S. \section{Background \label{sec: section3}} USMCA’s passage was touted as a “triumph…for workers everywhere across America” (\citeasnoun{Pramuk2020}). By including the HW-LVC requirements, USMCA committed “all parties to…encourage more production of automobiles and auto parts in the United States” (\citeasnoun{Finance2019}). The USTR stated that “the new rules of origin will achieve (the) goal” of “discourag(ing) the outsourcing of American automotive jobs, and instead encourage more investment and manufacturing jobs here in the United States” (\citeasnoun{USTR2019a}). The HW-LVC establish, for the first time in a trade agreement, a rules of origin (ROOs) requirement stipulating that a minimum of 40(45) percent labor value content for passenger cars (light trucks) be produced in a North American facility where the average production worker wage is a minimum of \$16 per hour (\citeasnoun{USTR2020a}). Under the standard staging regime, producers have three years to meet specified LVC thresholds that gradually increase over the period (table \ref{table:1}a). Producers also had the option to petition for an alternative staging regime with a longer transition period—five years—to ensure future production will meet the new LVC standards (table \ref{table:1}b) (\citeasnoun{FR2020b}). Alternative staging regime criteria for approval differed according to whether the vehicles covered under the petition constituted more than ten percent of the producer's total passenger vehicle or light truck production.\footnote{Thirteen companies’ petitions were approved. The companies are Cooperation Manufacturing Plant Aguascalientes (COMPAS), FCA North America Holdings LLC, Ford Motor Company Honda North America, Inc., Hyundai Motor America, Kia Motors Manufacturing Georgia, Kia Motors Mexico, Nissan North America Inc., Tesla Inc., Toyota Motor North America Inc., Volkswagen Group of America, Inc. Volvo Car Corporation (\citeasnoun{USTR2020b}). See table \ref{table:2} for data about the companies listed.} %Several auto manufacturers will be operating on an alternative staging schedule, as listed in table \ref{table:2}, but xx companies, comprising xx share of the Mexican autos industry, will be conforming to the standard staging schedule. Furthermore, % Between 2018 and 2019, numerous companies, including several of those whose petitions subsequently were approved, have stated either in public press statements or in conversations with USTR that USMCA will cause them to increase domestic auto investment (\citeasnoun{USTR2019b}). According to USTR, within five years, USMCA is estimated to support an additional 76, 000 automotive sector jobs. This figure, which would increase the domestic workforce by 7.6 percent, includes 22,800 assembly jobs, as detailed by specific information or public announcements made by automakers (\citeasnoun{USTR2019b}). \section{Empirical Strategy and Data \label{sec: section4}} We investigate how the welfare of specific worker subgroups fares during the HW-LVC anticipatory window spanning publication to entry into force. The HW-LVC are novel (no other U.S. trade agreement has them) and new (they were first made public on September 30, 2018). While their novelty presents no barrier to robust analysis---the economic effects of both price floors and trade liberalization have a vast literature---their newness severely limits post-implementation data availability. Frictions in the labor market and multi-year phase in periods for both the standard and alternative staging schedules call for informed decision making in construction of the identification strategy. In the next section, we discuss our identification strategy, the data used, and the details of our multistep econometric approach. \subsection{Identification} To determine the anticipatory effect of USMCA’s HW-LVCs on U.S. auto workers, we begin by identifying workers associated with any of four auto related manufacturing industry groups, as classified under the North American Industry Classification System (NAICS).\footnote{NAICS 3361 (motor vehicle manufacturing), 3362 (motor vehicle trailer manufacturing), 3363 (motor vehicle parts manufacturing), and 3369 (other transportation equipment manufacturing). (\citeasnoun{Census2020})} These auto workers are considered treated. Then, under denoted specifications of our model, we disaggregate treatment cohort status by gender and race. As discussed in section 2, under NAFTA, female workers disproportionately fared worse than their male counterparts, and Blacks were over-represented among workers certified by USDOL to have been negatively impacted by NAFTA. Under specification A, we classify female workers as treated and male workers as untreated. Under specification B, we classify Black workers as treated and non-Black workers as untreated. We then proceed to establish a treatment start date. The earliest possible start date is September 30, 2018, the date when the first draft of USMCA was made public. Numerous later dates are also viable. On May 17, 2019 the U.S. removed the Section 232 steel and aluminum tariffs on Canada and Mexico, thus eliminating a diplomatic hurdle to USMCA’s passage in Congress (\citeasnoun{Law3602020}). On December 13, 2019, the Agreement passed the House. It passed the Senate on January 16, 2020, was signed by the President on January 29, 2020, and entered into force on July 1, 2020. The standard regime phase-in period will be complete on July 1, 2023, and the alternate regime phase-in period is scheduled to be completed on July 1, 2025. A case could be made for choosing any of these dates. We choose September 30, 2018 as our starting point for a treatment date. The HW-LVC’s publication eliminates information asymmetry for industry stakeholders. Indeed, in the months following the announcement of the published ROOs, six automakers publicly announced domestic investments totaling 15.4 billion, due, in part, to the need to comply with USMCA's ROOs (\citeasnoun{USTR2019b}). We balance the need to capture agile industry reaction to the HW-LVCs with the acknowledgment that the labor market responds slowly to trade shocks, and that results generated from an aggressive treatment start date might suffer from underestimation (\citeasnoun{ArtucETAL2010}). Even so, we see October 1, 2018 as a viable start date. Because certain indicators are surveyed monthly, and others are surveyed yearly, we synchronize the start date of monthly and yearly samples of our data and choose March 2019 as our treatment start date. Our choice of a treatment start date that predates the fullness of liberalization is in line with Hakobyan and McLaren (2016), which uses data from the 2000 decennial census as the treatment date for assessing NAFTA’s local labor market effects (although NAFTA was only fully liberalized in 2008) (\citeasnoun{HakobyanMcLaren2016}). Cross-sectional microeconomic data on worker demographics is sourced from the monthly Current Population Survey (CPS), where we retain workers age 16-64 for whom industry of employment is reported at the North American Industry Classification System (NAICS) 4-digit industry group code level. We control for sex, “race”, education, state of residence, age, union membership, and veteran status. \subsection{Model} To determine the effect of the HW-LVC's announcement on the hours and wages of U.S. auto workers, we begin by estimating a baseline, two-period equation in which the outcome variable is either average hours worked or natural log wage. We restrict our sample to U.S. based auto workers, as explained in the identification section. Under model 1, the outcome variable $Y_it$ represents either hours worked at an individual’s “main job” in the previous week or worker’s total wage and salary income from the previous calendar year. The variable “average hours worked” is collected in the monthly CPS sample. The variable “wage and salary income” is collected in the yearly CPS Annual Social and Economic Supplement (ASEC) sample. $Post_t$ is an indicator variable with a value of “1” if a worker was interviewed on or after the treatment start date of March 2020. Under the baseline specification, the coefficient of interest precedes the variable $Post_t$. $X_i$ is a vector of demographics measuring age, education, race and sex. We implement increasing levels of controls by implementing union membership and state of residence, denoted $FE_i$. \begin{equation}\label{eq:1} Y_{it} = \theta_{i}+\gamma_{1}Post_{t}+\gamma_{2}'X_{i}+\gamma_{3}'FE_{i}+\epsilon_{it} \footnote{Baseline Equation: (Two-Period Difference)} \end{equation} We extend the two-period baseline specification to incorporate a second difference, namely whether the auto industry worker is employed in a production occupation. In so doing, we hone our focus on the group of workers directly impacted by the HW-LVC, which stipulates that a given fraction of an automobile must be produced in a facility where the average production wage is \$16/hr or higher (table \ref{table:1}a and b). Under (2), $Production_i$ is an indicator variable with a value of “1” denoting workers in production occupations. Under this difference-in-differences specification, the coefficient of interest precedes the variable ($Post*Production_{it}$), denoting the treatment group of production workers surveyed during the post-HW-LVC-announcement period. The remainder of the variables and coefficients from (1) retain their denotations. \begin{equation}\label{eq:2} Y_{it} = \alpha_{i}+\beta_{1}(Post*Production_{it})+\beta_{2}Post_{t}+\beta_{3}Production_{i}+\beta'X_{i}+\beta_{5}'FE_{i}+\epsilon_{it} \footnote{Baseline Equation (Difference-in-Differences: Production Workers)} \end{equation} After establishing baseline estimates for the HW-LVC effect on all automotive industry workers (equation 1) and on production workers (equation 2), we move on to estimating whether traditionally fragile subgroups—women and Blacks—exhibit a disproportionate welfare response to the HW-LVC announcement. We employ the following differences-in-differences econometric strategy, with two specifications: \begin{equation}\label{eq:3} Y_{it} = \alpha_{i}+\beta_{1}(Post*Subgroup_{it})+\beta_{2}Post_{t}+\beta_{3}Subgroup_{i}+\beta'X_{i}+\beta_{5}'FE_{i}+\epsilon_{it} \footnote{Difference-in-Differences (Subgroup Analyses)} \end{equation} Under specification A, female workers in the post-announcement period comprise the subgroup of interest. Under specification B, Blacks in the post-announcement period are treated. Under this model, we estimate the effect of the HW-LVC on both the full sample of autoworkers and the subsample of workers who identify as associated with a production occupation. Under model 3, $Y_{it}$ represents hours worked or total wage and salary income from the previous calendar year at an individual’s primary employment in the previous week. $Subgroup_{i}$ is an indicator variable with a value of “1” denoting workers who are part of the treatment subgroup according to specification (females under specification A and Blacks under specification B). The remainder of the variables from previous equations retain their denotations. Lastly, we use a triple differences model to estimate the effect of the HW-LVC announcement on production workers surveyed in the post-announcement period who are members of one of subgroups of interest. The coefficient of interest precedes the treatment group variable $Post_{t}*Subgroup_{i}*ProdWorker_{i}$, and all other variables from previous equations retain their denotations. \begin{equation}\label{eq:4} Y_{it} = \alpha_{i}+\beta_{1}(Post*Subgroup_{i}*ProdWorker_{i})+\beta_{n}D+\beta_{k}DD+\beta_{g}'X_{i}+\epsilon_{it} \footnote{Triple Differences (Production worker and Subgroup Analyses). In the term $\beta_{n}D$, n=3, and D denotes the single term FE variables $Post_{t}$, $Subgroup_{i}$, and $ProdWorker_{i}$. In the term $\beta_{k}D$, k=3, and DD denotes the two-variable interaction terms, $Post_{t}*Subgroup_i$, $Post_{t}*ProdWorker_{i}$, and $ProdWorker_{i}*Subgroup_{i}$} \end{equation} Under this model, equation (4), we use the full sample of workers. Altogether, we estimate the effect of the HW-LVC using four models, two specifications, and 2 samples. \subsection{Descriptive Statistics} A key assumption in difference-in-differences estimation is that treatment and control groups exhibit parallel trends prior to an intervention, such that post-period treatment group deviation from smooth cohort trends can be causally attributed to the intervention. Figures \ref{fig:2}, \ref{fig:3}, \ref{fig:4}, and \ref{fig:5} illustrate the pre-period trends for hours worked and wages. Wages trended similarly for groups separated by gender and race (figures \ref{fig:3} and \ref{fig:5}). Hours worked also trended similarly for both groups over time, with the exception of the fourth quarter of 2018, which exhibited a non-parallel uptick in hours worked for both women (figure \ref{fig:2}) and Blacks (figure \ref{fig:4}). In the pre-HW-LVC announcement period, the difference in mean hours worked between women and men was a statistically significant 1.72 hours per week, with women working fewer hours than men (Table 3). In addition, the wages for women were lower than for men (a difference of over \$13,000) For all covariates listed, with the exception of duration of unemployment, there was a statistically significant difference in levels in the pre-HW-LVC publication period. Between Blacks and Nonblacks, the pattern was similar to the pattern exhibited by females/males: Blacks worked fewer hours than Nonblacks in the pre-period (a statistically significant 0.88 hour difference) and had an average wage income lower than their Nonblack counterparts (a difference of approximately \$8.50) (table 4). One notable difference, however, was that within production occupations, Blacks were as likely as their Nonblack counterparts to hold supervisory positions. Between women and men, men had a statistically significant higher incidence of supervisory status in production occupations (table \ref{table:3}). \section{Results \label{sec: section5}} Table \ref{table:5} reports the results of the effect of the HW-LVC on all autoworkers generally, using baseline equations (1) and (2). We use Current Population Survey data from 2016 to 2020. Columns (1) – (3) have increasing levels of control: column (1) implements demographic controls, including education, veteran status, age, gender, and Black/Nonblack. Column (2) includes both demographic controls and controls for union membership. Households of union members historically have boasted a 10-20\% higher family income than non-union households, with returns that are, on average, even higher for Nonwhites (\citeasnoun{FarberETAL2021}). Column (3) includes demographic controls and union membership, and adds state fixed effects. By implementing state fixed effects, we control for state-specific macroeconomic trends, such as unemployment rate, minimum wage, and “right to work” status. Under the baseline specification, estimated with the full set of demographic controls, union status and state-of-residence fixed-effects, the outcome “average hours worked” remain statistically unchanged in the post-HW-LVC-publishing period both for auto workers generally (table \ref{table:5}, full sample, column 3) and for production autoworkers specifically (table \ref{table:5}, full sample, column 4). During the same period, the wages of the full sample of workers increased. The coefficient (0.064***) in column (3) of table 5 represents a 6.6\%, statistically significant increase in wages for the general autoworker population.\footnote{Percent change under a log-linear model with a categorical independent variable and log dependent variable is calculated as $\ 100(e^{\beta_{1}}-1)$, where $\beta_{1}$ is the coefficient on $Post_{t}*Subgroup_{i}$ of equation (1a) or (1b).} Production workers, however, see no such pay bump: their coefficient of -0.096 is negative and greater in magnitude of the general population’s coefficient, though it is not statistically significant. Thus, when compared to their wages in the pre-announcement period and the wages of non-production occupation autoworkers from both periods, production worker wages post-announcement show no improvement. The production worker subsample limits analysis of the effect of the HW-LVC to production occupation autoworkers, the group whose wages are utilized in HW-LVC calculations to determine the rules of origin. In contrast to the full subsample of auto industry workers, production occupation autoworkers do not earn a wage premium in the post-announcement period; across the increasing levels of controls (table 5, production workers subsample, columns 1-3), coefficients on wage remain small in magnitude and not statistically significant. Thus, when compared to their wages in the pre-announcement period, production worker wages post-announcement show no improvement. A reduction in average hours worked might be a contributing factor to production workers’ lackluster wage results. Though the estimates for average hours worked are negative and not significant in the full sample, in the production worker subsample, the estimates are larger in magnitude and weakly significant under the full set of controls, indicating a 1.1\% decline in hours worked in the post period (table \ref{table:5}, production worker subsample, column 3). Given these baseline estimates—no change in wages for production workers, a slight decrease in hours worked for production workers, an increase in industry wages, and no change in industry hours worked—we turn to comparing the estimates for our subgroups to our baseline estimates. The difference-in-differences full autoworker sample estimates in table \ref{table:6} are the gendered corollary for the full sample estimates for table \ref{table:5}, columns 1-3. The difference-in-differences production worker sample estimates in table \ref{table:6} are the gendered corollary for table \ref{table:5}. The full sample estimates under table \ref{table:6} indicate that post-announcement female autoworker wages and hours were not statistically distinct from the wages and hours of females in the pre-period or males in either period. In addition, post-announcement female production worker hours were not statistically distinct from the wages and hours of female production workers in the pre-period or male production workers in either period. Female production workers wages, however, are substantially different from the control group, comprised of female pre-period peers and male peers of either period. Female production workers’ coefficient of (-0.243**) (table \ref{table:6}) indicates that in the post period, female production workers earned 21.6\% less than their peers. \footnote{The two-period baseline specification's 6.6\% change in wages denotes wage growth relative to the pre-announcement period. The difference-in-differences estimate for female production workers denotes a change in wages for post-announcement female production workers relative to their peers (males from both periods and pre-announcement female production workers, and as such, might not indicate wage decline for the treatment group. For example, if peer wages rose and post-announcement female production worker wages remained unchanged, the relative change could still be negative.} Thus, the HW-LVC publishing is associated with a slight bump in overall wages for autoworkers, but a dramatic decline in wages for post-announcement period female production autoworkers. Next, we juxtapose the effect of the HW-LVC on Black workers, as denoted in table \ref{table:7}. The difference-in-differences full autoworker sample estimates in table \ref{table:7} are the ``race'' disaggregated corollary for the full sample estimates for table \ref{table:5}, columns 1-3. The difference-in-differences production worker sample estimates in table \ref{table:7} are the race disaggregated corollary for table \ref{table:5}. The first notable distinction between tables \ref{table:5}, \ref{table:6}, and \ref{table:7} is that the coefficient on hours per week is largest in the Black full worker subsample. The estimate falls outside the traditional bounds of statistical certainty, but could be a facet influencing the larger positive (yet still not significant) coefficient on wages for the full sample of autoworkers. The second notable distinction among the tables is that Black workers have no statistically significant results. This lack of significance could be statistical power issue, given that more detailed demographic analysis necessarily lowers the number of treated observations. Nevertheless, among production workers, Black post-period workers—whether or not in production occupations--fared no worse statistically than the control group comprised of their Nonblack peers from both periods and their Black peers from the pre-period. Last, we consider the triple difference estimators, which quantify the association of the HW-LVC with a worker’s status as a female (Black) post-period production worker versus the field. The tiny and statistically insignificant triple differences estimates, reported in table \ref{table:8}, for post-period female production worker hours and post-period Black production worker wages affirm our previous statements that these indicators trended similarly to aggregated baseline. Though the coefficient on the triple differences Black post-period production worker hours worked estimate is larger in magnitude, the large standard error, combined with trivial size and insignificance of the difference-in-differences estimate in table \ref{table:7}, suggests that Black production workers are not driving the estimate. As in table 6, we see that female post-period production workers fared worse than their peers; the coefficient (-0.438***) in column (3) of table 8 represents a 35.5\%, statistically significant decrease in wages for the post-period female production autoworker. We delve further to investigate the source of the dramatic change in wages for post-period female production workers, especially given that Black production worker wages and hours, and female production worker hours, trended similarly to the larger, aggregated sample. Figures \ref{fig:6} and \ref{fig:7} picture the kernel density estimates for female and male production workers, respectively, over our period of study.\footnote{The kernel density estimator is a nonparametric means of estimating the probability density function of a given variable.} For male production workers, the post-period density plot in figure \ref{fig:7} has shifted right; while the pre-period plot has a pronounced mass at lower wages (less than \$10,000 dollar range), the post-period plot has a pronounced mass at higher wages (\$40,000 to \$60,000). In contrast, for female production workers, the post-period density plot in Figure 4a has become more peaked relative to the pre-period plot. Notably, the post-period peak centers around the value \$33,600 (the dashed green line in the plot), which is the yearly wage income that corresponds with an hourly wage of \$16/hour.\footnote{We compute a back-of-the-envelope yearly income figure of \$33,600 by multiplying 50 weeks*42 hours per week (given the average hours worked from tables \ref{table:3} and \ref{table:4})*\$16/hour.} As contrasted with the pre-period female plot, which has mass at lower wages (\$0 to \$20,000 dollar range) and higher wages (\$50,000 to \$90,000), the post-period plot has less mass at those aforementioned wage ranges, and a pronounced mass at the more central wages of (\$30,000 to \$50,000). This distributional evidence suggests higher wage female production jobs loss as a possible mechanism for the reduction in wages associated with the publishing of the HW-LVC rules. Last, we investigate industry level events around the time of the HW-LVC publishing to shed more light on the marked decrease in female wages. On November 26, 2018, approximately 2 months after the HW-LVC’s publishing, GM announced that future products would be allocated to fewer plants in 2019, and that five North American plants would be “unallocated” in 2019, leaving thousands of workers at those plants unemployed.\footnote{The unallocated plants were located in Michigan (Detroit-Hamtramck Assembly in Detroit and Warren Transmission Operations in Warren, Michigan); Canada (Oshawa Assembly in Oshawa, Ontario, Canada); Ohio (Lordstown Assembly in Warren, Ohio.); and Maryland (Baltimore Operations in White Marsh, Maryland.) (\citeasnoun{GM2018}) It was reported that some workers employed at plants slated for idling will be offered jobs at other plants, but these job will require long-distance moves. (\citeasnoun{LawrenceHall2018})} We regressed the natural log wage/average hours worked on a triple differences interaction term equal to one for post-period female(Black) production workers living in the metro areas associated with the idled plants. As in Table \ref{table:6}, where we report estimates for the nationwide sample of female production workers, the results for average hours worked for female production workers in the GM closure metro areas estimated under the triple differences specification is positive, but not significant (table \ref{table:8}, column 4).\footnote{The coefficient on the triple interaction term is 2.041, standard error is 2.104.} At -1.229, the estimate for natural log wage under the triple differences specification, while not statistically significant, is negative and large, representing a 70.7\% decrease in wages for treated individuals. In addition, the average hours estimate for Black production workers in the region is very large and highly significant, representing a 15.9\% reduction in hours worked for Blacks living in the plant closure region. This result for Black hours is curious, especially when paired with the coefficient on wages that continues to be small and insignificant. \section{Robustness \label{sec: section6}} As a robustness check, we run a placebo regression to assess the validity of the difference-in-differences estimator. We estimate the relationship between and hours worked/wages and gender over all specifications in the years 2010 to 2016, a period predating widespread political discussion of NAFTA's renegotiation. We estimate the placebo regression using the progressive controls implemented in tables \ref{table:6} and \ref{table:7}, and, as reported in table \ref{table:9}, no gendered divergence from smooth cohort trends. \section{Conclusion \label{sec: section7}} We examine the wages and employment of U.S. auto workers to see if publication of the high wage labor value content rules (HW-LVC), announced in September 2018, exerted a disparate effect on female and Black workers. Multiyear phase-in periods for enforcement of LVC requirements could result in gradual labor responses from Mexican auto manufacturers (and correspondingly attenuated follow-on responses from American auto manufacturers), but anecdotal evidence suggests that upon publication, some automotive companies moved quickly to articulate plans for dealing with the \$16 average production wage stipulated under the USMCA rules of origin. Generally, we find that announcement of the HW-LVC had no statistically significant effect on the wages or hours of Black or female workers. The exception was for female production workers, who experienced a 21.6\% decline in wages relative to their peers in the period subsequent to the rules’ publishing. Our results indicate that, at least in the short term before the HW-LVS's implementation, U.S. female production workers have yet to experience the rules' welfare benefits. As these trends were established before entry into force, empirical analyses of HW-LVC implementation should incorporate these anticipatory effects. %%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 1 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \newpage \begin{figure}[h!] \centering \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure1.pdf}}{This figure graphs the evolution of U.S. direct investment abroad in the Mexican transportation industry on a historical-cost basis. The data are presented in millions of U.S. dollars. There is a vertical line at 1994, indicating the start of the NAFTA period.} \caption{U.S. Direct Investment Position Abroad on a Historical-Cost Basis (Source: Bureau of Economic Analysis, Balance of Payments and Direct Investment Position Data, accessed 2/3/21)} \label{fig:1} \end{figure} \newpage \begin{comment} Alt Text: This figure graphs the evolution of U.S. direct investment abroad in the Mexican transportation industry on a historical-cost basis. The data are presented in millions of U.S. dollars. There is a vertical line at 1994, indicating the start of the NAFTA period. \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 2 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \newpage \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure2.pdf}}{This figure separately graphs the average hours worked per week by men and women in the U.S. automobile industry. The graph presents data by quarter from Q1 2016 to Q4 2019.} \caption{Female and male trends in average hours worked per week by quarter} \centering \label{fig:2} \end{figure} \newpage %%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 3 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \newpage \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure3.pdf}}{This figure separately graphs the average wage earned per week by men and women in the U.S. automobile industry. The graph presents data by year from 2016 to 2019.} \caption{Female and male trends in average wages earned per week by year} \centering \label{fig:3} \end{figure} \newpage %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 4 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \newpage \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure4.pdf}}{This figure separately graphs the average hours worked per week by black and non-black workers in the U.S. automobile industry. The graph presents data by quarter from Q1 2016 to Q4 2019.} \caption{Black and non-black trends in average hours worked per week by quarter} \centering \label{fig:4} \end{figure} \newpage %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 5 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure5.pdf}}{This line graph for auto workers by race shows that the wages for Black and Nonblack workers trended similarly in the pre-announcement period from 2016 to 2019 .} \centering \caption{Black and non-black trends in average wages earned per week by year} \label{fig:5} \end{figure} \newpage \begin{comment} Alt Text: This figure separately graphs the average wage earned per week by black and non-black workers in the U.S. automobile industry. The graph presents data by year from 2016 to 2019. \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 6 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure6.pdf}}{This kernel density estimate graph for female workers by period shows that the plot for male worker in the post-announcement period became more peaked in the center of the distribution, at the value corresponding to an annual salary of 16 dollar per hour. The greatest decreases in mass in the post period came in the wage ranges from 10,000 to 15,000 dollars, 40,000 to 60,000 dollars, and 70,000 to 90,000 dollars.} \centering \caption{Estimated Kernel Density of Wage and Salary Earnings of Female Production Workers} \label{fig:6} \end{figure} \newpage \begin{comment} Alt Text: This figure graphs the estimated kernel density of wage and salary earnings of female production workers in the automotive industry, before and after USMCA. \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Figure 7 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{figure}[h!] \pdftooltip{\includegraphics[page=1, width=\linewidth]{Figure7.pdf}}{This kernel density estimate graph for male workers by period shows that the plot for male worker in the post-announcement period displayed a rightward shift. The greatest increase in mass in the post period came in the wage range from 40,000 to 80,000 dollars.} \centering \caption{Estimated Kernel Density of Wage and Salary Earnings of Male Production Workers} \label{fig:7} \end{figure} \newpage \begin{comment} Alt Text: This figure graphs the estimated kernel density of wage and salary earnings of male production workers in the automotive industry, before and after USMCA. \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 1 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table1.pdf} \centering \caption{Labor Value Content Requirements} \label{table:1} \end{table} \newpage \begin{comment} Alt Text: [Tables 1a and 1b detail the labor value content standard staging regimes for both passenger vehicles (table 1a) and light trucks (table 1b). The table columns are date of entry into force, LVC percent requirement, high wage material and manufacturing expenditures percent, high wage technology expenditures percent, and high wage assembly expenditures percent. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 2 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table2.pdf} \centering \caption{Companies with approved alternative staging plans} \label{table:2} \footnote{“NA” denotes unavailable information. List of companies is from \citeasnoun{USTR2020b}. Company information is from Bureau van Dijk, accessed January 31, 2021. [Latest operating revenue, global ultimate owner, and number of employees for companies]. https://orbis.bvdinfo.com/.} \end{table} \newpage \begin{comment} Alt Text: [Table 2 lists companies with approved alternative staging plans. The table columns are company name, data date, number of employees (as of the data date) parent company/global ultimate owner, and company operating revenue. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 3 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table3.pdf} \centering \caption{Male/Female Pre-Period Difference of Means} \label{table:3} \end{table} \newpage \begin{comment} Alt Text: [ Table 3 is a difference of means table by gender. The columns are female mean, male mean, mean difference, and P-value. The rows are average hours worked, duration of unemployment, age, education, production occupation, production supervisors, natural log wage, union membership, and Black. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 4 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table4.pdf} \centering \caption{Black/Non-Black Pre-Period Difference of Means} \label{table:4} \end{table} \newpage \begin{comment} Alt Text: [ Table 4 is a difference of means table by race. The columns are Black mean, Nonblack mean, mean difference, and P-value. The rows are average hours worked, duration of unemployment, age, education, production occupation, production supervisors, natural log wage, union membership, and female. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 5 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table5.pdf} \centering \caption{Baseline Estimates (Post-period autoworkers are “treated”)} \label{table:5} \end{table} \newpage \begin{comment} Alt Text: [Table 5 presents baseline model estimates of the effect of the HW-LVC announcement on the hours per week and wages of both the full sample of autoworkers and the production autoworker subsample. The estimates in column 1 use demographic controls, the estimates in column 2 use both demographic and union membership controls, and those in column 3 use demographic controls, a union membership control, and state fixed effect. Column 4 presents a difference-in-differences estimate with production workers in the post-announcement period as the treatment group and all others as the control group, and uses the full set of controls. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 6 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table6.pdf} \centering \caption{Difference-in-Differences Estimates Under Specification A (Females are “treated”)} \label{table:6} \end{table} \newpage \begin{comment} Alt Text: [ Table 6 presents difference-in-differences model estimates of the effect of the HW-LVC announcement on the hours per week and wages of both the full sample of autoworkers and the production autoworker subsample. Under this framework, female workers are considered treated. The estimates in column 1 use demographic controls, the estimates in column 2 use both demographic and union membership controls, and those in column 3 use demographic controls, a union membership control, and state fixed effect. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 7 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table7.pdf} \centering \caption{Difference-in-Differences Estimates Under Specification B (Blacks are “treated”)} \label{table:7} \end{table} \newpage \begin{comment} Alt Text: [Table 7 presents difference-in-differences model estimates of the effect of the HW-LVC announcement on the hours per week and wages of both the full sample of autoworkers and the production autoworker subsample. Under this framework, Black workers are considered treated. The estimates in column 1 use demographic controls, the estimates in column 2 use both demographic and union membership controls, and those in column 3 use demographic controls, a union membership control, and state fixed effect. ] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 8 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table8.pdf} \centering \caption{Triple Differences Estimates (Post-period production workers in specific subgroups are “treated”)} \label{table:8} \end{table} \newpage \begin{comment} Alt Text: [ Table 8 presents triple differences model estimates of the effect of the HW-LVC announcement on the hours per week and wages of both the full sample of autoworkers. Under specification A, female post-announcement production workers are considered treated. Under specification B, Black post-announcement production workers are considered treated. The estimates in column 1 use demographic controls, the estimates in column 2 use both demographic and union membership controls, and those in column 3 use demographic controls, a union membership control, and state fixed effect. Column 4 controls for whether the respondent lived in the region affected by numerous GM plant closures, along with the aforementioned controls] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Table 9 here % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \begin{table}[h!] \includegraphics[page=1, width=\linewidth]{Table9.pdf} \centering \caption{Robustness Estimates for Difference-in Differences Estimates Under Specification A (Females are “treated”)} \label{table:9} \end{table} \newpage \begin{comment} Alt Text: [ Table 9 presents robustness checks for hours per week and natural log wage estimates for female workers. Samples covered include the full sample of workers, the production worker subsample, and production workers in the GM plant closure region subsample. The estimates in column 1 use demographic controls, the estimates in column 2 use both demographic and union membership controls, and those in column 3 use demographic controls, a union membership control, and state fixed effect. No coefficients in table 9 are statistically significant, indicating that the estimates from tables 5-8 are robust.] \end{comment} %%%%%%%%%%%%%%%%%%%%%%%%%%%% % Bibliography % %%%%%%%%%%%%%%%%%%%%%%%%%%%% \newpage \singlespacing \bibliographystyle{dcu} \bibliography{biblio} \newpage \end{document} @Article{Aguayo-TellezETAL2012, Title = {{Did Trade Liberalization Help Women? The Case of Mexico in the 1990s}}, Author = {Aguayo-Tellez, Ernesto and Airola, Jim and Juhn, Chinhui}, Journal = {NBER Working Paper}, Publisher = {National Bureau of Economic Research}, Year = {2012}, Number = {16195} } @Article{Aguayo-TellezETAL2013, Title = {{Trade Liberalization and Gender Inequality}}, Author = {Aguayo-Tellez, Ernesto and Airola, Jim and Juhn, Chinhui}, Journal = {The American Economic Review}, Year = {2013}, Volume = {103}, Number = {3}, Pages = {269-73} } @Article{ArtucETAL2010, Title = {{Trade Shocks and Labor Adjustment: A Structural Empirical Approach}}, Author = {Artuc, Erhan and Chaudhuri, Shubham and McLaren, John}, Journal = {American Economic Review}, Year = {2010}, Volume = {100}, Number = {3}, Pages = {1008-45} } @Article{Benguria2020, Title = {The Impact of NAFTA on U.S. Local Labor Market Employment}, Author = {Benguria, Felipe}, Journal = {SSRN}, Year = {2020} } @Article{BLS2021, Title = {{“Automotive Industry: Employment, Earnings, and Hours—National Average Hourly Earnings, not Seasonally Adjusted}}, Author = {{U.S. Bureau of Labor Statistics}}, Year = {2021}, URL = {https://www.bls.gov/iag/tgs/iagauto.htm} } @Article{BurfisherETAL2001, Title = {{The Impact of NAFTA on the United States}}, Author = {Burfisher, Mary E. and Robinson, Sherman and Thierfelder, Karen}, Journal = {Journal of Economic Perspectives}, Year = {2001}, Volume = {15}, Number = {1}, Pages = {125-144} } @Article{CaliendoParro2015, Title = {{Estimates of the Trade and Welfare Effects of NAFTA}}, Author = {Caliendo, Lorenzo and Fernando Parro}, Journal = {The Review of Economic Studies}, Year = {2015}, Volume = {Vol. 82}, Number = {No. 1}, Pages = {1-44} } @Article{CBO1993, Title = {{A Budgetary and Economic Analysis of the North American Free Trade Agreement}}, Author = {{Congressional Budget Office}}, Journal = {A CBO Study}, Year = {1993}, URL = {https://www.cbo.gov/sites/default/files/103rd-congress-1993-1994/reports/199307budgetaryanofnafta.pdf} } @misc{Census2020, URL = {https://www.census.gov/naics/?input=31&chart=2017}, Author = {{U.S. Census Bureau}}, Year = {2020}, Note = {Accessed: 2-5-2021}, Title = {{North American Industry Classification System (NAICS)}} } @Article{ChatzkyETAL2020, Title = {{NAFTA and the USMCA: Weighing the Impact of North American Trade}}, Author = {Chatzky, Andrew and McBride, James and Aly Sergie, Mohammed}, Journal = {{Council on Foreign Relations - Backgrounder}}, Year = {2020} } @Article{CruzRiker2014, Title = {{The Impact of NAFTA on U.S. Labor Markets}}, Author = {De La Cruz, Justino and David Riker}, Journal = {USITC Working Paper Series}, Year = {2014}, Volume = {2014-06-A} } @Article{CuevasLopez2019, Title = {{The effects of NAFTA on economic growth}}, Author = {Cuevas, Victor and Lopez, Roger}, Journal = {{Investigación Económica}}, Year = {2019} } @Article{Enriquez2020, Title = {{Japanese Suppliers Choose to Raise Wages}}, Author = {Enriquez, Alejandro}, Journal = {Mexico Business}, Year = {2020} } @Article{FarberETAL2021, Title = {{Unions and Inequality over the Twentieth Century: New Evidence from Survey Data}}, Author = {Farber, Henry S. and Herbst, Daniel and Kuziemko, Ilyana and Naidu, Suresh}, Journal = {The Quarterly Journal of Economics}, Year = {2021}, Volume = {136}, Number = {3}, Pages = {1325–1385} } @Article{Finance2019, Title = {{Myths and Facts: USMCA}}, Author = {{United States Senate Committee on Finance}}, Journal = {Chairman's News}, Year = {2019}, URL = {https://www.finance.senate.gov/chairmans-news/myths-and-facts-usmca} } @Article{FR2020a, Title = {High-Wage Components of the Labor Value Content Requirements Under the United States-Mexico-Canada Agreement Implementation Act}, Author = {{Department of Labor}}, Journal = {Federal Register}, Year = {2020}, Volume = {85 FR 39782}, Pages = {39782-39817} } @Article{FR2020b, Title = {Procedures for the Submission of Petitions by North American Producers of Passenger Vehicles or Light Trucks To Use the Alternative Staging Regime for the USMCA Rules of Origin for Automotive Goods}, Author = {{Office of the U.S. Trade Representative}}, Journal = {Federal Register}, Year = {2020}, Volume = {85 FR 22238}, Pages = {22238-22244} } @Article{FrancisZheng2011, Title = {{Trade Liberalization, Unemployment and Adjustment: Evidence from NAFTA Using State Level Data}}, Author = {Francis, John and Zheng, Yuqing}, Journal = {Applied Economics}, Year = {2011}, Volume = {Vol. 43}, Pages = {1657-1671} } @Article{FukaoETAL2003, Title = {{An Econometric Analysis of Trade Diversion under NAFTA}}, Author = {Fukao, Kyoji and Okubo, Toshihiro and Stern, Robert M.}, Journal = {North American Journal of Economics and Finance}, Year = {2003}, Volume = {14}, Number = {3}, Pages = {3-24} } @Article{Gertz2020, Title = {{Did Trump’s tariffs benefit American workers and national security?}}, Author = {Gertz, Geoffrey}, Journal = {Policy 2020 Brookings}, Publisher = {Brookings Institute}, Year = {2020}, URL = {https://www.brookings.edu/policy2020/votervital/did-trumps-tariffs-benefit-american-workers-and-national-security/} } @Article{GM2018, Title = {{General Motors Accelerates Transformation}}, Author = {{General Motors}}, Journal = {General Motors Press Release}, Year = {2018}, URL = {https://plants.gm.com/media/us/en/gm/news.detail.html/content/Pages/news/us/en/2018/nov/1126-gm.html} } @Article{GoldbergPavcnik2005, Title = {{Trade, Wages, and the Political Economy of Trade Protection: Evidence from the Colombian Trade Reforms}}, Author = {Goldberg, Pinelopi Koujianou and Pavcnik, Nina}, Journal = {Journal of International Economics}, Year = {2005}, Volume = {66}, Number = {1}, Pages = {75-105} } @Article{HakobyanMcLaren2016, Title = {{Looking for Local Labor Market Effects of NAFTA}}, Author = {Hakobyan, Shushanik and McLaren, John}, Journal = {Review of Economics and Statistic}, Year = {2016}, Volume = {98}, Number = {4}, Pages = {728–741} } @Article{HakobyanMcLaren2017, Title = {{NAFTA and the Gender Wage Gap}}, Author = {Hakobyan, Shushanik and McLaren, John}, Journal = {Upjohn Institute Working Paper}, Publisher = {W.E. Upjohn Institute for Employment Research}, Year = {2017}, Volume = {17-270} } @Book{HufbauerSchott2005, Title = {{NAFTA Revisited: Achievements and Challengess}}, Author = {Hufbauer, Gary Clyde and Schott, Jeffrey J.}, Publisher = {Peterson Institute for International Economics}, Year = {2005}, Number = {No. 332} } @misc{IPUMS2020, URL = {https://cps.ipums.org/cps-action/variables/IND#description_section}, Author = {Ruggles, Steven and Flood, Sarah and Goeken, Ronald and Grover, Josiah and Meyer, Erin and Pacas, Jose and Sobek, Matthew}, Year = {2020}, Note = {Accessed: 2-5-2021}, Title = {{IPUMS USA: Version 10.0}} } @Article{LawrenceHall2018, Title = {{GM decision to close Detroit-Hamtramck met with shock, anger}}, Author = {Lawrence, Eric D. and Hall, Christina}, Journal = {Detroit Free Press}, Year = {2018}, URL = {https://www.freep.com/story/money/cars/general-motors/2018/11/26/gm-cut-jobs-detroit-hamtramck/2113462002/} } @Article{Kim2020, Title = {{Carmakers Choose Mexico Over U.S., Despite Trump's Claim NAFTA Would Create Jobs}}, Author = {Kim, Soo}, Journal = {Newsweek}, Year = {2020} } @Article{KlierRubenstein2017, Title = {{Mexico’s Growing Role in the Auto Industry Under NAFTA: Who Makes What and What Goes Where}}, Author = {Klier, Thomas and Rubenstein, James}, Journal = {Economic Perspectives}, Publisher = {Federal Reserve Bank of Chicago}, Year = {2017}, Volume = {Vol. 41}, Number = {No. 6}, Pages = {49-68} } @Article{KoopmanETAL2010, Title = {{Give Credit Where Credit is Due: Tracing Value Added in Global Production Chains}}, Author = {Koopman, Robert and Powers, William and Wang, Zhi and Wei, Shang-Jin}, Journal = {{NBER Working Paper}}, Publisher = {National Bureau of Economic Research}, Year = {2010}, Number = {16426} } @Article{Law3602020, Title = {{Reports On USMCA Offer Industry-Specific Guidance For Cos}}, Author = {Law360}, Year = {2019} } @Article{MarangerSwiecki2016, Title = {{The Growing Role of Mexico in the North American Automotive Industry: Trends, Drivers and Forecasts}}, Author = {Maranger Menk, Debbie and Swiecki, Bernard }, Journal = {Center for Automotive Research White Paper}, Publisher = {Center for Automotive Research}, Year = {2016} } @Article{Peterson2014, Title = {{NAFTA 20 Years Later}}, Author = {Cimino-Isaacs, Cathleen and Hufbauer, Gary Clyde, and et al.}, Publisher = {Peterson Institute for International Economics}, Year = {2014}, Volume = {PIIE Briefing No. 14-3}, Pages = {304-311} } @Article{Pramuk2020, Title = {{House Democrats and the White House have a deal to move forward with USMCA trade agreement}}, Author = {Pramuk, Jacob}, Journal = {CNBC.com}, Year = {2020}, URL = {https://www.cnbc.com/2019/12/10/house-democrats-and-trump-administration-reach-usmca-trade-deal.html} } @Article{Public2021, Title = {{Trade Discrimination: The Disproportionate, Underreported Damage to U.S. Black and Latino Workers From U.S. Trade Policies}}, Author = {Public Citizen Global Trade Watch}, Journal = {International Journal of Business & Economics Perspectives}, Year = {2021} } @Article{Roma2007, Title = {{NAFTA’s and CUSFTA’s Impact on International Trade}}, Author = {Romalis, John}, Journal = {Review of Economics and Statistics}, Year = {2007}, Volume = {89}, Number = {3}, Pages = {416-435} } @Article{Salvador2018, Title = {{President Trump’s claims about NAFTA and the U.S. auto industry}}, Author = {Rizzo, Salvador}, Journal = {The Washington Post}, Year = {2018} } @Article{SaureZoabi2014, Title = {{International trade, the gender wage gap and female labor force participation}}, Author = {Saure, Philip and Zoabi, Hosny}, Journal = {Journal of Development Economics}, Year = {2014}, Volume = {Vol. 111}, Pages = {17-33} } @Article{Schwartz2016, Title = {{Carrier Workers See Costs, Not Benefits, of Global Trade}}, Author = {Schwartz, Nelson}, Journal = {The New York Times}, Year = {2016} } @Article{Shaiken1993, Title = {{Myths About Mexican Workers}}, Author = {Shaiken, Harley}, Publisher = {Democratic Study Center}, Year = {1993} } @Article{Townsend2007, Title = {{Do Tariff Reductions Affect the Wages of Workers in Protected Industries? Evidence from the Canada-U.S. Free Trade Agreement}}, Author = {Townsend, John}, Journal = {The Canadian Journal of Economics }, Year = {2020}, Volume = {40}, Number = {1}, Pages = {69-92} } @Article{Tyson2017, Title = {{Americans generally positive about NAFTA, but most Republicans say it benefits Mexico more than U.S}}, Author = {Tyson, Alex}, Journal = {The Pew Fact Tank}, Publisher = {Pew Research Institute}, Year = {2017} } @Article{USITC2016, Title = {{Economic Impact of Trade Agreements Implemented Under Trade Authorities Procedures}}, Author = {{U.S. International Trade Commission}}, Year = {2016}, Volume = {4614} } @Article{USTR1993, Title = {{North American Free Trade Agreement}}, author = {{Office of the U.S. Trade Representative}}, Year = {1993}, URL = {https://ustr.gov/about-us/policy-offices/press-office/ustr-archives/north-american-free-trade-agreement-nafta} } @Article{USTR2019a, Title = {{USMCA Estimated to Support 76,000 Additional American Auto Jobs, Billions in New Auto Manufacturing Investments}}, author = {{Office of the U.S. Trade Representative}}, Year = {2019} } @Article{USTR2019b, Title = {{ESTIMATED IMPACT OF THE UNITED STATES-MEXICO-CANADA AGREEMENT (USMCA) ON THE U.S. AUTOMOTIVE SECTOR}}, author = {{Office of the U.S. Trade Representative}}, Year = {2019} } @Article{USTR2020a, Title = {{United States—Mexico—Canada Trade Fact Sheet Modernizing NAFTA into a 21st Century Trade Agreement}}, author = {{Office of the U.S. Trade Representative}}, Year = {2020} } @incollection{USTR2020b, title = {{Alternative Staging}}, booktitle = {{United States-Mexico-Canada Agreement}}, author = {{Office of the U.S. Trade Representative}}, Year = {2020}, URL = {https://ustr.gov/trade-agreements/free-trade-agreements/united-states-mexico-canada-agreement/alternative-staging} } @Article{VillarealFergusson2017, Title = {{The North American Free Trade Agreement (NAFTA)}}, Author = {Villareal, M. Angeles and Fergusson, Ian F.}, Journal = {Congressional Research Service}, Year = {2017}, URL = {https://fas.org/sgp/crs/row/R42965.pdf} } @Article{WenReinbold2019, Title = {{The U.S. Auto Labor Market since NAFTA}}, Author = {Wen, Yi and Reinbold, Brian}, Journal = {On the Economy}, Publisher = {Federal Reserve Bank of St. Louis}, Year = {2019} } @Article{WolduETAL2018, Title = {{An Empirical Analysis of the Impact of Nafta on Economic Growth of Its Member Countries}}, Author = {Woldu, Habte G. and Alborz, Shawn and Myneni, Nitish}, Journal = {International Journal of Business & Economics Perspectives}, Year = {2018}, Volume = {Vol. 13}, Number = {No. 1}, Pages = {1-17} }