The Intangible Gender Gap: An Asset Channel of Inequality

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 1322 August 2021 The Intangible Gender Gap: An Asset Channel of Inequality Carlos F. Avenancio-Leon and Leslie Sheng Shen Please cite this paper as: Avenancio-Leon, Carlos F. and Leslie Sheng Shen (2021). “The Intangible Gender Gap: An Asset Channel of Inequality,” International Finance Discussion Papers 1322. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/IFDP.2021.1322. NOTE: International Finance Discussion Papers (IFDPs) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the International Finance Discussion Papers Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at www.ssrn.com.

The Intangible Gender Gap: An Asset Channel of Inequality∗ Carlos F. Avenancio-Le´on† Leslie Sheng Shen‡ May 2021 Abstract We propose an “asset channel of inequality” that contributes to gender inequities. We establish thatindustrieswithlow(high)genderpaygapshavehigh(low)sharesoftangibleassets. Because asset tangibility determines firms’ ability to collateralize assets and borrow, credit conditions affect industries differently. We show that credit expansions further reduce the pay gap in lowpay-gap industries while leaving it unaffected in high-pay-gap industries, making low-pay-gap industries more appealing for women. Consequently, gender sorting across industries increases, which then cements gender roles and accentuates workplace gender bias. Ultimately, credit expansions help women “swim upstream” but also reinforce glass ceilings. JEL Classification: J71, O16 Keywords: Gender Pay Gap, Credit Markets, Asset Tangibility, Equitable Finance ∗We are grateful to David Autor, Sapnoti Eswar, Janet Gao, Eitan Goldman, Nandini Gupta, Isaac Hacamo, Craig Holden, Hilary Hoynes, Margaret Jacobson, Rupal Kamdar, Preetesh Kantak, Kristoph Kleiner, Nirupama Kulkarni, Eben Lazarus, Jordan Matel, Christopher Palmer, Mar´ıa Cecilia P´erez, Vincenzo Pezone, Alessio Piccolo, Leonardo Rafael, Batchimeg Sambalaibat, Julien Sauvagnat, Petra Vokat´a, Zhenyu Wang, Frank Warnock, Michal ZatorandDayinZhangforcommentsandsuggestions. SeminarparticipantsattheAEAAnnualMeetings,theMIT Finance Lunch, the IU–Bloomington Finance Internal Seminar, and the Midwest Finance Association Meeting also providedusefulfeedback. TammySunjuLeeprovidedexcellentresearchassistance. Theviewsinthispaperaresolely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or any other person associated with the Federal Reserve System. †University of California at San Diego. Email: cavenancioleon@ucsd.edu ‡Federal Reserve Board. Email: leslie.shen@frb.gov.

I. Introduction Changes in financial conditions—credit contractions and expansions—are frequent1 and have been shown to affect income inequality (Beck, Levine, and Levkov 2010, Philippon and Reshef 2012). But how do financial conditions affect group inequality? To explore this question, we focus on the connection between financial conditions and gender inequality. We propose an “asset channel of inequality” and show that it, in particular, drives the persistence of gender inequities. When financial conditions change, differences in firms’ ability to collateralize assets restructure the set of productive investments held by the firm. This restructuring propels changes in labor composition and demand both across industries and, due to preexisting gender imbalances, across gender lines. When these changes in labor benefit women at the bottom of the pay distribution but not at the top,changesinfinancialconditionsmaysimultaneouslyhelpwomen“swimupstream”andreinforce the glass ceiling. In this paper, we document these dynamics by showing how financial deregulation reduced the genderpaygapatthebottomofthepaydistributionwhilepropellinggendersortingoutofthetop. We establish that industries with high gender pay gaps have higher wages and less tangible assets; conversely, industrieswithlowgenderpaygapshavelowerwagesandmoretangibleassets. Because asset tangibility determines firms’ collateral and ability to borrow, and thus choice of project and ensuing labor demand, financial deregulation has different effects on workers from industries with different levels of tangibility. In more equitable industries (i.e., industries with a lower pay gap and moretangibleassets), firmsincreaseborrowingandtakeonmorepositiveNPVprojectsinresponse to financial deregulation, which increases their demand for labor. In more inequitable industries (i.e., industries with a higher pay gap and more intangible assets), firms do not significantly change theirborrowingbutmustrelinquishmarginallypositiveNPVprojectsinresponsetoincreasedlabor competitionfromthemoreequitableindustries, whichlowerstheirdemandforlabor. Differencesin labordemandbetweenequitable industries andinequitableindustries, togetherwithhigherrelative pay for women in the more equitable industries, lead to gender sorting between the more equitable and inequitable industries. We further show that this sorting accentuates workplace gender bias by cementing gender roles. Our findings contribute to the understanding of the persistence and evolution of the gender 1. Over the past few decades, prominent instances of financial shocks include the waves of financial deregulation in the United States, the 2008 financial crisis, potential COVID-19 credit crunch, etc. 1

pay gap. Gender inequities in the labor market have been large and persistent. While pay for men and women has narrowed, especially during the 1980s, women still earn, on average, 20 percent less than men (Blau & Kahn, 2017). In addition, the narrowing of the gender pay gap took place at the bottom and center of the wage distribution rather than at the top, and progress slowed afterwards (Blau & Kahn 1997, 2017). Our findings suggest that the waves of financial deregulation in the 1980s, by differentially affecting wages by industry and inducing gender sorting across industries, contributed to the bottom-up narrowing of the gender pay gap, and by cementing gender roles, made gender inequities persistent. Ouranalysisproceedsinthreesteps. First,weexploitvariationininterstateandintrastateU.S. bank deregulation to estimate the effects of credit expansions along the gender pay gap distribution across industries. Because pay for men and women converged significantly during the 1980s, as we noted above, we fix pay gap levels by industry prior to 1980 and categorize industries into high pay gap, medium pay gap, and low pay gap based on their preexisting pay gap levels. We find that whilerelativewagesforwomendidnotchangeinthehigh-pay-gapindustriesinresponsetobanking deregulation, they increased by 5% in the low-pay-gap industries, controlling for Mincerian traits (education, experience, and experience squared). These results are robust to alternative methods of industry categorization. Second, we explore the role of assets in linking changes in credit conditions to the gender pay gap. We begin by documenting the industry-level relationship between gender pay gaps and asset tangibility: The share of tangible assets is consistently and materially lower in the high-pay-gap industries than in the low-pay-gap industries. This differentially affects the ability of the two types of industries to post assets as collateral. We then show that asset tangibility funnels changes in credit conditions by differentially affecting firm borrowing, investment decisions, and demand for workers across the two types of industries. More specifically, in response to bank deregulation, the low-pay-gap (or high-asset-tangibility) industries increased borrowing, further shifted their asset compositiontowardstangibleassets, andincreaseddemandforlabor. Incontrast, thehigh-pay-gap (or low-asset-tangibility) industries did not change borrowing behavior, further shifted their asset composition towards intangible assets as evidence by increased R&D expenditures, and lowered labor demand. Changes across all these dimensions point to a restructuring of the labor market. This restructuring of the labor market altered the dynamics of gender through the interplay of two main forces. One, the low-pay-gap industries, despite being more equitable, are on average low paying, while the high-pay-gap industries are high paying. Two, surpluses in the high-pay-gap 2

industries, owing to low female representation in these industries, are disproportionately accrued to male workers.2 Moreover, these rents make men relatively more likely to transition into the high-pay-gap industries. In contrast, the low-pay-gap industries compete for workers with low risk of transitioning into the higher-paying high-pay-gap industries, which are more likely to be women. Thiscompetitioninducesthelow-pay-gapindustriestoincreaserelativepayforwomen,whichleads to higher women participation in these industries as a byproduct. Increases in the relative pay in the lower-paying (but more equitable) jobs combined with low femalerepresentationinthehighpayingjobsalterstheopportunitycostforwomenrelativetomen. Thiscreatesincentivesforwomentoselectintomoreequitablebutlower-payingindustries, or, conversely, abstain from participating in less equitable but higher-paying ones. Indeed, we document that following bank deregulation, women are more likely to stay in the low-pay-gap industries and exit the better-paying high-pay-gap industries. This sorting behavior leads to a persistence of the gender pay gap by perpetuating gender imbalances across industries. Furthermore, higher participation in lower-paying industries makes women more vulnerable to economic downturns: We show that credit contractions disproportionately reduce women’s wages in the low-pay-gap industries, reverting the gains from credit expansions.3 Lastly,weshowthatthisassetchannelofinequalityhasdownstreamimplications: theresulting differences in gender sorting may cement gender roles in the long run. Individuals, both male and female, may interpret the differences in sorting and the resulting gender imbalance through gendered lens and conclude that women are less suitable for some jobs, or that it is less important for women to pursue a career, or that women have a comparative advantage for staying at home. We directly test for changes in gender norms fo this sort by analyzing how bank deregulation, through industrial composition, affects measures of sexism derived from the General Social Survey (GSS) data. We find that, following credit expansion, attitudes toward women in the workplace worsened. In particular, attitudes toward women of men and individuals with children worsened more. On net, gains in relative pay for women in the lower-paying industries offset losses arising from gender sorting (i.e., women sorting into lower-paying industries) at the extensive margin, leading 2. Thisisconsistentwithempiricalfactsdocumentedintherecentliterature,includingCard,Cardoso,andKline (2016) and Barth, Kerr, and Olivetti (2017). It is also consistent with the empirical findings documenting a positive relationshipbetweeninnovativeinvestmentsandrentsharingwithworkers(VanReenen1996)andthatrentscanbe disproportionately shared with male workers (Black & Strahan 2001; Kline et al. 2018). 3. See Appendix E for an analysis of the effects of credit contraction on the gender pay gap. 3

to an overall reduction in the gender pay gap. Nevertheless, the reduction comes at the cost of increased sorting that polarizes gender imbalances across industries, and changes in gender norms thatreflectsuchpolarization. Thistransformationingenderinequities–ratherthananunqualified declineinthepaygap–mayhavebeenacontributingfactortotheslowprogressinpayconvergence between women and men after the 1980s. Contribution to the Literature This paper furthers the understanding of the determinants of gender inequities and, in particular, the role of financial conditions on the evolution of gender inequities in the labor market. As such, this paper contributes to several lines of research in labor economics and finance. First, we contribute to the research on the factors affecting the persistent gap in pay between genders. Previousstudiesinthisareaemphasizeoneofseveralgeneralhypotheses: lackoftemporal flexibility in the structure of jobs in the labor market (Goldin 2014), cultural differences that translate into differences in choices (Goldin 2006), and gender differences in bargaining power (Babcock and Laschever 2003). We propose an alternative channel that complements these mechanisms and highlight how gender inequities can transition from wage gaps into inequities in job allocation. Second, while previous literature has documented the differences in earnings between women and men over the life cycle (Barth Et Al. 2017; Goldin Et Al. 2017), the determinants that explain therelationshipbetweengendersorting(intoparticularfirms, occupations, orindustries)andlower pay are less understood. One approach to assess this relationship is to evaluate whether external conditions force women to sort into lower paying firms (e.g. flexible hours, Goldin 2014; home production, Albanesi and Olivetti 2009). Nevertheless, there is also evidence showing that job pay decreases synchronously with increased female participation (Levanon et al. 2009). This finding suggests that approaching the relationship between pay and gender sorting from the perspective of the employer may be informative: why do female-dominated firms become lower paying? Our contribution addresses this question by showing how credit conditions can exacerbate differences in pay across industries, and then how those differences in pay may lead to sorting across gender lines and accentuation of gender norms. Third, we contribute to the literature on the real effects of financial liberalization4 by showing that credit liberalization propels changes in labor demand across industries and in the size of the 4. See,forexample,KingandLevine(1993),Demirguc-KuntandMaksimovic(1998),RajanandZingales(1998), BeckandLevine(2004),JayaratneandStrahan(1996),CetorelliandStrahan(2006),andBeck,Levine,andLevkov (2010). 4

gender pay gap within industries in a way that can influence industry gender composition and gender norms. In this regard, the paper closest to ours is Black and Strahan (2001), which shows thatgendergapsnarrowwithinthebankingsectorfollowingbankingderegulationasrentsavailable for sharing with workers decrease with competition. However, our paper differs considerably from the existing literature as it focuses on the gendered labor dynamics triggered by differences in collateral when financial conditions change. Fourth, by documenting an asset (as collateral) channel that drives inequality, we contribute to the literature on equitable finance that attempts to dissect the financial mechanisms that lead to economic redistribution (e.g., Beck, Levine, and Levkov 2010). II. Data & Methodological Approach Our goal is to evaluate how changes in credit conditions alter the cross-industry dynamics of the gender pay gap, and in particular, whether they differentially affected the pay gaps of industries by their (ex-ante) equitability. In other words, do changes in credit amplify preexisting equitability or inequitability within industries? To capture exogenous changes in credit conditions across industries, we exploit the temporal and spatial variation in U.S. bank deregulation. In this section, we first provide a brief background on U.S. bank deregulation. We then discuss our data and empirical approach, including how we categorize industries by their preexisting equitability. II.1 Intrastate and Interstate Banking Deregulation The events of U.S. bank deregulation during the 1970s–90s are well-documented, starting with Jayaratne and Strahan (1996). In sum, there were two main sets of deregulation events in the banking industry. The first one was the removal of restrictions on branching within states, which mostly occurred between 1970 and 1994. In line with the literature, we refer to this event as intrastate bank deregulation or simply branch deregulation. The other deregulation event was the removal of restrictions on cross-state ownership of banks.5 Following the lead of Maine, all states expect Hawaii started allowing entry of out-of-state bank holding companies with legislative changes taking place from 1978 to 1992. As in the literature, we henceforth refer to this event as interstate bank deregulation. 5. TheDouglasAmendmenttothe1956BankHoldingCompanyActeffectivelyprohibitedbankholdingcompanies from acquiring banks outside the state(s) where their headquarter(s) resided, unless states actively allowed the acquisitions. 5

Jayaratne and Strahan (1996) and Kroszner and Strahan (1999) provide a detailed analysis of thepoliticalandeconomicreasonsfortheexacttimingofthederegulationevents, pointingoutthat states did not deregulate their banks in anticipation of future good growth prospects. Additionally, studies have shown that bank deregulation led to increased competition among lenders and an improvement in the efficiency of the banking industry, which helped to facilitate firm borrowing and investment by relaxing financial constraint (Jayaratne and Strahan 1996, Black and Strahan 2002, Rice and Strahan 2010, Jiang et al. 2020). We therefore exploit the cross-state, cross-time exogenous variations in credit available to firms from banking deregulation to examine the causal effects of the relaxation in credit constraints on the cross-industry dynamics of the gender pay gap. II.2 Data Our main data comes from the March Supplement of the Current Population Survey (CPS) for the years 1976–2014.6 We restrict our sample to working-age full-time full-year workers in the private sector. To ensure that our estimates are not driven by industrial organization changes within the finance industry (Black and Strahan 2001), we exclude individuals working in the Finance, Insurance, and Real Estate (FIRE) industries. Our primary outcome variable of interest is individual hourly wage.7 The CPS also contains individual demographic information such as race, gender, age, and educational attainment, as well as detailed information on employment, including occupation and industry, prior year occupation and industry, type of employer (public vs. private), and county and state of work. The CPS incorporates probability sampling weights for each individual, which indicate their representativeness in the population. We use these sampling weights in all our specifications. We supplement the CPS data with the Compustat data to evaluate the effects of bank deregulation on firm borrowing, investment (including tangible asset and R&D spending), and measures ofprofitabilityperemployee(toassessefficientuseoflabor). WealsousetheGSSdatatoconstruct indexes of sexism following Charles, Guryan, and Pan (2018), which are used to evaluate the effects of bank deregulation on changes in gender norms in Section V.8 6. We start the analysis in 1976 because, in the CPS data, states can only be identified separately starting in the 1977 survey (which cover data from 1976), similar to Beck, Levine, and Levkov (2010). In Appendix J.1, as a robustness, we conduct our analysis using an expanded dataset that starts in 1968. 7. We use the log transformation of this outcome as our dependent variable. 8. In Appendix Section E, we also evaluate vulnerability of women to contractions in credit, using data from the FDIC call reports on mergers. 6

II.3 Empirical Specification & Industry Equitability Categorization To estimate the causal effect of changes in credit conditions on the cross-industry dynamics of the gender pay gap, we employ a generalized difference-in-differences design, exploiting cross-state, cross-year variation in the timing of intrastate and interstate banking deregulation. Specifically, we estimate the differential labor market outcomes for female workers relative to male workers across industries of varying preexisting equitability, in response to banking deregulation. To proxy for each industry’s preexisting equitability, we categorize industries by their pay gap levels in the first five years of CPS data (1976–1980), which is prior to our estimation sample period.9 We categorize industries into high-, medium-, and low-pay-gap industries based on the preperiod pay gap distribution using the 1990 Census Industry Codes (CIC). The high-pay-gap industries are defined as those in the top quartile of distribution; the low-pay-gap industries are those in the bottom quartile; the medium-pay-gap industries are those in between. We discuss the stability of this categorization scheme in Subsection II.4. Let Ω = {High,Medium,Low} denote the high-, medium-, and low-pay-gap industries, and Ik is a dummy variable indicating whether industry j falls into classification k ∈ Ω. Our primary j empirical specification takes the following form: (cid:88) (cid:88) (cid:88) Y = α+ β D ×Ik + γ D ×Ik ×F + δ Ik ×F (1) ijst k st j k st j i k j i k∈Ω k∈Ω k∈Ω (cid:88) + ζ Ik +πX +τ +µ +(cid:15) k j ijst t,female s,female ijst k∈Ω where D is a dummy denoting whether deregulation has taken place in state s and year t, F st i indicates whether individual i is female, X is a vector of demographic controls including Minijst cerian traits (education, experience, experience squared), race, and marital status, and τ t,female and µ are time-gender and state-gender fixed effects, respectively. Single order F term is s,female i absorbed by fixed effects. II.4 Stability of Industry Equitability Categorization Our empirical analysis embeds the assumption that the rank of industries by equitability is stable prior to 1980. We conduct four tests to examine the stability of the equitability categorization. First, because legislative changes leading to interstate deregulation took place after our categorization period (1976–1980) for all states except Maine (which took place in 1978), we repeat our 9. The estimation sample spans the years 1981 to 2014. The preperiod choice is driven by both data restrictions and the importance of the 1980s decade for understanding the evolution of the pay gap (Blau and Kahn 1997). 7

interstate deregulation analysis excluding Maine and find that our choice of industry categorization is stable (see Appendix F). Second, we note that intrastate deregulation changes occurred before our categorization period for 17 states, which raises the possibility that our industry categorization iscontaminatedbyintrastatederegulation(Amel1993). Inlightofthisconcern, weconductourintrastate deregulation analysis excluding the 17 states. We show that our categorization of industry equitability is not sensitive to this exclusion, which confirms that our industry categorization is not a result of our treatment (see Appendix F). Third, we show that alternative categorization methods, including categorization using the 1968–1972 CPS data, yield the same results (see Appendix J.) Fourth, in all our subsequent analyses, we present results using both interstate and intrastate bank deregulation and show that the estimates are virtually identical. II.5 Summary Statistics Employment Summary Statistics. In Table (1.A), we present summary statistics on characteristics of male and female workers across all industries (columns 1–2) and in the low- and high-pay-gap industries separately (columns 3–6). Hourly wage. Hourly wages are $5.43 lower for women than for men in the high-pay-gap industries on average, while the difference is only $0.99 in the low-pay-gap industries. This translates into a difference of −33 and −8.5% in hourly wage between women and men in the high- and low-pay-gap industries, respectively. Overall, women earn $3 dollars (22%) less than men for each hour of work. Demographics. Years of education for women are similar between the high- and low-pay-gap industries, at around 13.4 years of schooling. Male workers in the high-pay-gap industries have an additional 1.4 years of schooling on average. Age of workers is similar across industries and across genders, ranging from 39.7 to 40.9 years. Men tend to have 0.6–0.7 more years of experience than women across all industries. Labor force participation. Female labor force participation is visibly higher in the low-paygap industries at 41.9%, while the high-pay-gap industries have a higher female participation rate (38.3%) than the average rate (34.9%) in the full sample (which also contains the medium-pay-gap industries). The differences in female participation between the low- and high-pay-gap industries are stable over time, as shown in Panel A of Appendix Figure (C.5). Fraction of hourly-paid positions. The low-pay-gap and high-pay-gap industries exhibit visible differences in their employment of hourly-paid positions, as shown in Panel B of Appendix Figure 8

(C.5). The low-pay-gap industries employ a higher fraction of hourly-paid positions (around 61% of their total employees), and the share decreased in the high-pay-gap industries between 1990 and 2014 (from 50% to 40%).10 Moreover, the hourly-paid positions are held mostly by women in both industries, and the decline in hourly-paid positions in the high-pay-gap industries is attributed to men. Routine and nonroutine occupations. The two types of industries also exhibit different occupational needs. Using the occupational task measure developed by Autor, Levy, and Murnane (2003) fromtheDictionary of Occupational Titles (DOT),weclassifyanoccupationasroutine,nonroutine cognitive, or nonroutine manual. As shown in Panel A of Appendix Figure (C.3), the high-pay-gap industries rely more on workers to perform nonroutine cognitive tasks, and their reliance increased steeply over time (left panel). In contrast, the low-pay-gap industries employ a higher fraction of workers to perform nonroutine manual tasks, and the share such workers employed by the highpay-gap industries steadily drops over time from more than 20% in 1980 to near 10% in 2014 (right panel). In addition, there are no significant differences between the low and high-pay-gap industry in their reliance on routine jobs for either manual or cognitive tasks, as shown in Panel B of Appendix Figure (C.3). The share of routine task jobs declined in both industries over the sample period, with a more notable decreasing trend in the high-pay-gap industries. The declining trend may be attributed to the rise in computer technology, as routine tasks are more vulnerable to substitution by computers (Autor, Levy, and Murnane 2003). Overall, Appendix Figure (C.3) shows that the main difference in occupational needs between the low- and high-pay-gap industries lies in their reliance on workers performing different types of nonroutine tasks. Firm Summary Statistics. In Table (1.B), we present summary statistics on characteristics of public firms across all industries (column 1) and in the low- and high-pay-gap industries (column 2 and 3, respectively). Compared to firms in the high-pay-gap industries, firms in the low-pay-gap industries have slightly higher assets (a statistically insignificant difference of 3%), more workers, and higher revenues and income by worker. The high-pay-gap industries have lower book leverage (48 vs. 54%), higher Tobin’s Q (1.09 vs. 0.92), and lower levels of tangibility (0.22 vs. 0.55) than the low-pay-gap industries. Wefindthatthelow-pay-gapindustriesaremorereliantonexternalfinancingandmorecapital 10. Weplotthefractionofhourly-paidpositionsbyindustryonlyduringtheperiod1990–2014becauseofconstraint due to data availability. 9

intense than the high-pay-gap industries. We compute debt-to-asset ratios (for secured debt, debt notes, and long-term debts) and leverage by industries in Table (2). The low-pay-gap industries are consistentlymorereliantondebtthanthehigh-pay-gapindustriesregardlessofthedebtinstrument: the former industries are twice as likely to use secured debt, debt notes, and long-term debt than the latter industries, and leverage in the low-pay-gap industries is 7.4% higher. Furthermore, the low-pay-gap industries are more capital intense than the high-pay-gap industries throughout our estimation period. In Figure (1), we plot total assets, total plant and equipment, and total tangibilityperemployeebyindustries. Regardlessofinstruments,thelow-pay-gapindustriesexhibit higher capital intensity than the high-pay-gap industries throughout the sample period. The difference in reliance on external financing and capital as well as in employment needs across industries serve as important grounds for divergent industrial response following bank deregulation, which eventually leads to a restructuring of the labor market. Industry Summary Statistics: Frontier Industries and the Pay Gap. Table (3.A) lists industries exhibiting the highest and lowest pay gaps (Panel A) and the fastest and slowest growth (Panel B). Overall, service-oriented industries exhibit the highest pay gaps, which include Legal services, Advertising, Accounting services, Physicians, and Dentists. Agricultural and Care industries exhibit more equitable pay. Pay gaps in Physicians and Dentists offices are mostly driven by high levels of occupational segregation, where women dominate care taking activities like nursing. Table (3.B) shows that the industries exhibiting the highest levels of growth in employment from 1980 to 2000 are the high-pay-gap industries. Among the top 10 fastest-growing industries, seven are in the high-pay-gap category (including Computer and Data Processing Systems and Management and Public Relations Services), and only one is in the low-pay-gap category (Agricultural Chemicals). In contrast, the slowest-growing industries exhibit no obvious differences in their pay gap level: four industries exhibit low pay gaps while three exhibit high pay gap levels. It is also the case that the high-pay-gap industries typically pay more than the low-pay-gap industries throughout the sample period (Figure 2.A). Compared to the low-pay-gap industries, average pay is around 21% higher in the high-pay-gap industries. Furthermore, that difference is driven almost exclusively by higher wages for men (Figure 2.B). Combined with observations from Table (3.B), the data suggest that women benefited less from industry growth. Relative to the low-pay-gap industries, the high-pay-gap industries exhibit significantly higher growth in both employment and pay during the sample period. 10

III. Credit Conditions and Gender Inequality Table(4)presentstheestimationresultsonthedifferentialeffectsofbankingderegulationonwages by gender in the low-pay-gap and high-pay-gap industries based on Equation (1).11 Columns (1)– (4) show results on the effects of intrastate deregulation, and columns (5)–(8) show those on the effects of interstate deregulation. All specifications control for Mincerian traits (education, experience, experience squared), marital status, race, state-gender dummies and year-gender dummies. Columns (2)–(4) and (6)–(8) additionally control for age-gender dummies. Our results show significant heterogeneity in the effects of banking deregulation on wages across industries and gender. While the overall wage across industries declined by 4% in response to (intrastate and interstate) banking deregulation on average, wage in the high-pay-gap industries increased by 4% more than the average. In other words, wage for workers in the high-pay-gap industries increased by 8% relative to the medium- and low-pay-gap industries in response to the change in credit condition. The overall decline in wages can be explained by the changes in labor composition in response to the changes in credit conditions. As we had discussed in Section II.5, both the high-pay-gap and low-pay-gap industries relied less on routine workers following deregulation. In Appendix Table (H.8), we show that the overall decline in wage is driven by routine workers, consistent with the idea that nonroutine occupations are substituting routine occupations. Zooming into the effects of banking deregulation on wages by gender across industries, we find significant differential effects between the low-pay-gap and high-pay-gap industries. In the low-pay-gap industries, relative wages for women increased by about 5% in response to banking deregulation, as the decrease in wages accrues mostly to men. As shown in Figure (3) Panel A, this relativeincreasetookplaceimmediatelyafterderegulation. Incontrast,femaleworkersexperienced no relative increase in their pay in the high-pay-gap industries. These results indicate that bank 11. InAppendixL,wevalidatetherobustnessoftheestimationresultsusingaunstaggereddifference-in-differences design, in light of the concern that estimating heterogeneous treatment effects in a standard (staggered) differencein-differences design might negatively weight some events (Sun and Abraham 2020). We show that our results are similar when aligning events by event-time instead of calender-time. We provide more details in Appendix L. 11

deregulation improves wage parity between men and women only in lower-paying jobs.12 Thesedifferentialeffectsinpaybygenderaccruetopre-bankingderegulationwagedifferentials between low- and high-pay-gap industries by gender. Prior to banking deregulation, average pay for men is 21% higher in high-pay-gap than low-pay-gap industries, controlling for education and experience. The corresponding difference for women is only 7%. A similar stylized fact is observed in Figure (2) Panel B. When we incorporate the effects of bank deregulation, the difference in wage between high-pay-gap than low-pay-gap industries amplifies to 29% for men and 11% for women. Balance & Pre-Trends. Standard assumptions for difference-in-differences regressions require no significant difference between the treatment and control group and no pre-trends. To test these assumptions, we conduct a number of analyses, and show that there is balance—observable characteristics are similar between the treatment and control groups and the results are not driven by unobservable characteristics—and that there is not a pre-trend. First, we study the differences in characteristics between the states that are about to undergo bankderegulation(treatmentgroup)andthestateswherederegulationhadnotpassedandwillnot pass in the following year (control group) on average, to examine whether the two groups approximate an “apples-to-apples” comparison. Appendix Figure (B.1) Panel A illustrates the differences in the case of intrastate deregulation, and Panel B illustrates those for interstate deregulation. Evidently, most characteristics between the two groups of states are not different at the five percent level on average; the differences are economically small in magnitude and precise. The characteristic that varies the most between the two groups is the percentage of workforce that is black. States deregulating intrastate branching have 0.6% more black workers on average than the non-deregulating states (the average share of black workers in the deregulating states is 6%), while states deregulating interstate branching have 0.006% fewer black workers (average share is 7.5%). Nevertheless, both estimates are highly imprecise. Also, the percentage of nonroutine manual workers is marginally different between deregulating and non-deregulating states with a difference of 1.5% for intrastate deregulation (average share in the deregulating states is 26.5%) 12. While our results demonstrate convergence in pay in the low-pay-gap industries, it is not clear whether this convergence leads to an unambiguous reduction in the gender pay gap. To this end, we perform two simple Oaxaca- Blinder decompositions, as shown in Appendix Section D. We find that gains in relative pay for women in the lower-payingindustriesoffsetanylossesarisingfromgendersorting(i.e.,womensortingintolower-payingindustries) attheextensivemargin,leadingtoanoverallreductioninthegenderpaygap. However,asweshowinthesubsequent sections, the reduction comes at the cost of increased sorting that polarizes gender imbalances across industries and cement gender roles. 12

and 2.1% for interstate deregulation (average share is 26.4%). Overall, this analysis shows that observable characteristics are similar between the treatment and control groups, which mitigates concerns about unobservable institutional differences confounding our estimation results. Next, we assess whether the parallel trends assumption holds in a couple ways. First, we examine differences in preperiod trend between the treatment and control group. Appendix Figure (B.2) illustrates the differences in average yearly trends of a wide range of characteristics between the states that are about to undergo bank deregulation and the states where deregulation had not passedandwillnotpassinthefollowingyear. Asbefore, PanelAshowstheestimatesforintrastate deregulation, and Panel B show those for interstate deregulation. All average trends between the two groups are not different at the five percent level. The differences are economically small in magnitude and precise, including percentage of black workers and percentage of nonroutine manual workers. This evidence supports the assumption of parallel trends. Another way to study the parallel trends assumption is observing the behavior of outcomes of interest around the deregulation years in an event study. As we had shown in Figure (3) Panel A, the relative wage increase took place immediately after deregulation without a pretrend. In the subsequent section, we conduct similar event studies for all the other outcomes we study and show that there are no pretrends. In addition, we plot the raw likelihood of working in the highand low-pay-gap industries by gender 10 years prior and 10 years after intrastate deregulation Figure (4). The likelihood is computed by assigning −1 to workers in the low-pay-gap industries, 1 to workers in the high-pay-gap industries, and 0 otherwise, and then taking the average of the indicators by gender in each period, using the CPS data. The plot shows that the sharp changes in labor participation occurred after the passage of deregulation with no leading trends. IV. The Role of Assets In this section, we show that asset tangibility serves as a key force driving the relationship between credit conditions and gender inequality. IV.1 Asset Tangibility and Gender Pay Gap: Overview and Hypotheses We first document a new stylized fact on the relationship between asset tangibility and gender pay gap. In Panel A of Figure (1), we plot asset tangibility per employee for the low-pay-gap and high-pay-gap industries. As shown, more equitable industries tend to have a significantly higher share of tangible assets. Conversely, more inequitable industries have a higher share of intangible 13

assets. This empirical pattern holds not only on a per worker basis but also as a share of total assets (Panel B of Table 1). Relatedly, we observe that the high-pay-gap industries have more physical assets and lower total asset on a per worker basis than the low-pay-gap industries (Panel B and C of Figure 1). Overall, the observed relationship between asset tangibility and gender pay gap suggests that the low-pay-gap industries are less capital intensive. How does asset tangibility drive the effects of banking deregulation on gender inequality? First, the tangibility of assets affects firm borrowing. Second, differences in firm borrowing lead to differential investment decisions and demand for workers. Third, imbalances in demand for workers across genders lead to differential effects on gender pay gap in the high-asset-tangibility and low-asset-tangibility industries. Asset Tangibility and Firm Borrowing. The tangibility of assets affects firm borrowing because high- and low-tangible assets differ in debt capacity. Williamson (1988) and Shleifer and Vishny (1992) stress the importance of asset redeployability, or asset’s potential for alternative uses, for debt capacity. In case default occurs, assets can be seized by creditors and redeployed, increasing the value available to creditors. This reasoning is particularly relevant for tangible assets. Tangible assets can sustain more external financing by increasing the value available to creditors when default occurs (Almeida & Campello 2007). On the other hand, intangible assets, which can be, for example, in the form of R&D or brand name, contain limited capacity for pledgeability as collateral, even though they can provide the firm with a competitive edge (Lev 2000). Giventhedifferencesinpledgeabilitybetweentangibleandintangibleassets,bankderegulation could lead to differential firm borrowing behavior between high-asset-tangibility and low-assettangibility industries. While bank deregulation increases access to credit in general, we conjecture thatithasagreatereffectonborrowinginindustrieswithmoretangibleassets(i.e., industrieswith a lower pay gap), as the higher pledgeability of tangible asset enhances borrowing capacity. On the other hand, borrowing in industries with more intangible assets (i.e., industries with a higher pay gap) may not be affected as much, as intangible assets are harder to post as collateral. Hypothesis 1: In response to bank deregulation, the low-pay-gap industries, which have a higher share of tangible assets, are more likely to increase firm borrowing, while borrowing in the high-paygap industries, which have a lower share of tangible assets, is likely to remain unaffected. 14

Investment and Labor Decisions. Differences in firm borrowing in response to changes in credit condition could have divergent effects on investment and labor decisions. The relaxation of financial constraints following bank deregulation allows firms to dig into an untapped pool of positive NPV projects and change the composition of the labor force. While bank deregulation relaxes financial constraints in general, it has a stronger impact on investment in tangible assets, as tangible assets can be collateralized.13 This investment is likely to increase relative demand for labor in industries in which financial constraints are relaxed more—the low-pay-gap industries with higher share of tangible assets prior to deregulation. Hypothesis 2a: The low-pay-gap industries, because of their higher share of tangible assets (which help to relax financial constraints more following bank deregulation), are more likely to increase investment in tangible assets and increase labor demand relative to high-pay-gap industries. The relative increase in labor demand in industries with more tangible assets exerts upward pressure on wages, which we observe for non-routine workers.14 For industries with more intangible assets, theincreaseinwageputsmildlypositiveNPVprojects(themarginalprojects)intonegative territory, which should prompt them to relinquish these projects—a downscale-to-quality mechanism. In this case, the average revenue per employee and the average wage in low-asset-tangibility industries should increase. Conversely, the average revenue per employee is expected to decline in high-asset-tangibility industries, as they take on more positive (but lower and more marginal) NPV projects.15 In search for higher NPV projects, the low-asset-tangibility industries are likely to substitute into more intangible investment, such as R&D, which would face less competition from high-asset-tangibility industries. Hypothesis 2b: In the high-pay-gap industries, which have more intangible assets, average revenue per employee is expected to increase, and firms are more likely to substitute into investing in R&D. In the low-pay-gap industries, the average revenue per employee is expected to decline, as they take 13. Forexample, inacarpurchase, thecaritselfcanbeusedasacollateral, whichrelaxesthefinancialconstraint for that investment. 14. Forroutineworkers,weobserveadeclineinwages,whichsuggeststhat,inaggregate,industriesaresubstituting away from routine workers. 15. The reason for the increase (decrease) in average revenue per employee in low-asset-tangibility (high-assettangibility) industries can be elucidated through a simple example. Consider two projects, Project 1 and Project 2. The revenue before wage for Project 1 and 2 is 100 and 50, respectively, and each requires one employee. If wage moves from 49 to 51, only the first project will be taken. The average revenue would increase from 75 to 100. This scenario corresponds to the expected effect in low-asset-tangibility industries in our context. Conversely, if a firm initiallyinvestsinthefirstprojectandwagegoesdownfrom51to49,itwilltakeonthesecondproject,whichdrags down its average revenue. This scenario corresponds to the expected effect in high-asset-tangibility industries in our context. 15

on more positive (but lower) NPV projects. From Labor Decisions to the Gender Pay Gap. How do these asset composition and labor decisions affect the gender pay gap? The effect stems from the optimal hiring decisions by the low-pay-gap (or high-asset-tangibility) industries whose demand for labor has increased because of banking deregulation. In particular, we argue that any potential gender imbalance in hiring decisions by the high-pay-gap (or low-asset-tangibility) industries would affect hiring in the lowpay-gap industries. In the following, we discuss the mechanisms through an illustrative example. In Appendix A, we formalize the argument in a model. Consider two types of employers: high-pay employers who can pay high wages (e.g., Amazon) and low-pay employers who can’t (e.g., the Washington Post), all else equal. In our context, the high-pay employers correspond to firms in the low-asset-tangibility or high-pay-gap industries, and the low-pay employers correspond to firms in the high-asset-tangibility or low-pay-gap industries: As shown in Figure (2a), wages in the high-pay-gap industries are 3% higher than those in the median-pay-gap industries during the sample period of 1980–2000, while wages in the low-paygap industries are 18% lower than the average wage. There are two potential gendered hiring approaches: employers who hire based on employee skill and, by optimizing skills, statistically discriminate against a particular group (statistical discrimination), and employers who hire employees strictly based on taste on gender (taste discrimination). When the low-pay employers look to hire an employee based on a specific skill, e.g., writing, it can choose between two sets of employees. Both sets have similar writing skills but one has superior computer science skills. The optimal choice for the low-pay employer is to hire the employee from the group with less adequate computer science skills. (Why? Otherwise, they would have to pay the reservation wage for a computer scientist.) If it is the case that there is a gender imbalance in the employee pool for high-pay employers, then the low-pay employer would necessarily have the reverse gender imbalance through a reverse statistical discrimination mechanism. An alternative gendered hiring practice is when the employers have particular taste about the gender of the employee. When the high-pay employer taste-discriminates against one gender, it is optimal for the low-pay employer to hire from the employee group that is taste-discriminated against—a reverse taste discrimination mechanism. We do not take a stance on whether high-pay employers statistically or taste discriminate. What we are arguing is that if either is the case, reverse discriminatory practices would be optimal for the low-pay employers. This argument is broadly related to Arrow (1973). 16

Basedonthesemechanisms,thelow-pay-gapindustriesinourcontextwouldhiremorefromthe employee group that is discriminated against by the high-pay-gap industries. If the high-pay-gap industries statistically or taste discriminate against women, then it is optimal for the low-pay-gap industry to hire women over men. This leads to an increased demand for women that puts upward pressure on women’s relative wages. Hypothesis 3: Following bank deregulation, the relative wage for women would increase in the low-pay-gap industries, as their labor demand for women increases relative to the high-pay-gap industries. IV.2 Firm Borrowing We first test whether banking deregulation differentially affected firm borrowing between the lowpay-gap (or high-asset-tangibility) and high-pay-gap (or low-asset-tangibility) industries (Hypothesis 1). Specifically, we examine the effects on firm overall debt growth, long-term debt growth, and debt ratio. Table (5) shows the results on the effects of bank deregulation on firm borrowing changes for industries that had higher or lower gender pay gap prior to deregulation, as specified in Equation (1). The results show that both debt and long-term debt increased in the low-pay-gap industries in response to deregulation. Specifically, intrastate deregulation increased overall debt and long-term debt growth by 5 log points in these industries. On the other hand, there was no significant growth in debt in the high-pay-gap industries, but their debt ratio declined, which suggests that non-debt financing increased in these industries. Overall, theresultsareconsistentwithourhypothesis. Thelow-pay-gapindustries, whichhave more tangible assets prior to deregulation, increase borrowing in response to deregulation and the resulting relaxation of financial constraints, as the higher pledgeability of tangible assets enhances their borrowing capacity. On the other hand, the high-pay-gap industries, which have more intangible assets prior to deregulation, do not significantly change their borrowing, as intangible assets are harder to post as collateral. IV.3 Firm Asset Composition and Labor Decisions Next we proceed to test whether banking deregulation differentially affected firm asset composition and labor decisions between the low-pay-gap and high-pay-gap industries. 17

Asset Composition. First, we examinehow deregulationaffectedtangibleassetcomposition, as measured by the share of tangible assets, in the two types of industries (Hypothesis 2a). Columns (1)–(3) in Table (6) show the results. Industries with low pay gap prior to deregulation significantly increased their relative investment in tangible assets. Specifically, intrastate and interstate deregulation increased their relative investment in tangible assets by 2 log points. On the other hand, the high-pay-gap industries did not significantly change their tangible asset composition. These results support our hypothesis. Banking deregulation is expected to relax financial constraints more for the low-pay-gap industries because they have more tangible assets prior to deregulation, which can be collateralized. Additional investment in tangible assets are easier to make because these assets can be collateralized as well. Therefore, the low-pay-gap industries increase their relative investment in tangible assets in response to deregulation. Labor Demand. This increase in relative investment by the low-pay-gap industries is likely to increase their relative demand for labor. To examine whether that is the case, we plot the difference in labor share between the high-pay-gap and low-pay-gap industries before and after banking deregulation, as illustrated by the solid black line in Figure (4). Recall that the high-paygap and low-pay-gap industries are categorized based on whether their pay gap falls in the top and bottom quartile, respectively, of the pay gap distribution from 1976 to 1980. In other words, by construction, the share of labor in the high-pay-gap and low-pay-gap industries each makes up 25% of total labor market at the period of construction. Thus, the difference in labor share between the high-pay-gap and low-pay-gap industries is roughly zero before deregulation, as shown by the solid black line. In the years after deregulation, the difference in labor share between the high-pay-gap and low-pay-gap industries turned negative, which indicates a change in labor demand towards to low-pay-gap industries and away from high-pay-gap industries. We complement this empirical finding on the change in labor demand at the extensive margin with within-firm estimations using Compustat data. Columns (1)–(3) of Table (7) show results from estimations of the differential effects of banking deregulation on firm employment between the high-pay-gap and low-pay-gap industries. Based on the estimates in column (3), employment in the low-pay-gap industries increased by 7 log points (relative to the omitted medium-pay-gap industries) in response to banking deregulation, controlling for firm and state-year fixed effects and firm controls. On the other hand, employment in the high-pay-gap industries decreased by 4 log points, although this estimate is not statistically significant. These estimates are robust to 18

alternative specifications (columns 1 and 2). Project Composition. Next, wetestwhetherbankingderegulationdifferentiallyaffectsaverage revenue per employee between the high-pay-gap and low-pay-gap industries (Hypothesis 2b). The increase in labor demand in the low-pay-gap industries exerts upward pressure on wage, which should change the composition of projects that are undertaken by both types of industries, as we explained in Section IV.1. We test for this effect, and the results are shown in columns (4)–(6) in Table (7). The results in column (6) indicate that relative revenue per employee increased by 13 log points in the high-pay-gap industries in response to banking regulation, controlling for firm and state-year fixed effects and firm controls. This result is consistent with the idea that high-pay-industries relinquish marginal projects when the marginal cost exceeds marginal revenue (downscale-to-quality), increasing the average revenue per employee. In contrast, relative revenue peremployeedeclinedinthelow-pay-gapindustriesinresponsetobankingregulation, asrelaxation of financial constraint allows them to undertake additional lower positive NPV projects. Based on the estimates in column (4), revenue per employee in the low-pay-gap industries declined by more than 12 log points relative to the medium-pay-gap industries following deregulation, which implies a total decline of 17 log points (as revenue per employee decreased by 5 log points on average across industries). Most of the relative decline goes away with the inclusion of firm controls (column 6), which implies that revenues per employee declined by at least 5 log points in the low-pay-gap industries, just as in the medium-pay-gap industries. Revenue per employee proxies for surplus absorbed by all stakeholders of the firm, including creditors, employees, and the employer itself. We devise an approach to decompose the total surplus into components absorbed by the employers, the employees, or others such as creditors. We first remove potential surplus absorbed by the creditors by dropping non-operating expenses from revenue. This corresponds to testing for the differential effects of banking deregulation on net income + operating expense per employee between the high-pay-gap and low-pay-gap industries. Net income captures the surplus absorbed by the employers and does not include wages, while operating expense is driven in large part by wages—the surplus absorbed by employees. Next, we focus on net income solely as the dependent variable, or surplus solely absorbed by employers. Columns(7)–(9)ofTable(7)showtheresultsonnetincome+operatingexpenseperemployee, andcolumns(10)–(12)showthoseonnetincomealone. Firstnoticethatestimatesusingnetincome + operating expense per employee as the dependent variable are similar to those using revenue per 19

employee. This suggests that the results on revenue per employee are not driven by changes in credit conditions such as interest rates. Based on the results in column (9), net income + operation expense increased by 13 log points in response to banking deregulation, which is the same as the estimate on revenue (column 6). At the same time, net come per employee increased by a lower amount (9 log points, based on column 12). This means that operating expenses, including wages, are absorbing part of the effects. The differences in outcomes between the two sets of results proxy thechangeinsurplusabsorbedbytheemployees. Takentogether, ourresultsshowthattherelative increase in revenues absorbed by workers in the high-pay-gap industries is around 4 log points. On the other hand, the net relative loss for the low-pay-gap industries is around −4 log points. We compare the estimates on the relative changes in revenue absorbed by workers to those on wages from Table (4). The two sets of estimates are of the same magnitude. As shown in Table (4), banking deregulation increased (decreased) the absolute wages in the high-pay-gap (low-pay-gap) industries by 4 log points, the same as our estimates on relative increase (decrease) in revenues absorbed by workers in the two industries. We also study the increase in absolute wages using an event study version of Eq. (1) following Borusyak & Jaravel (2017). As illustrated in Figure (3) Panel B, absolute wages in the high-pay-gap industries sharply increased after deregulation, while pre-trends are statistically indistinguishable from zero. Project Substitution. As the high-pay-gap industries relinquish marginal projects, it is likely that they substitute into more intangible investments such as R&D, where there is less labor competitionfromthelow-pay-gapindustries. Wetestthedifferentialeffectsofbankingderegulation on R&D investment for the low-pay-gap and high-pay-gap industries. The results are shown in columns (4)–(6) of Table (6). Based on the estimates in column (4), R&D spending in highgender-pay-gap industries increased by 31 log points more than other industries following banking deregulation. ThisdifferenceisdriveninpartbyareductioninR&Dspendinginlow-andmediumgender-pay-gap industries.16 The results are robust to the inclusion of state×year fixed effects and firm controls. 16. The reduction is consistent with the findings in Chava et al. (2013), who document a decline in innovation following intrastate deregulation. 20

IV.4 Gender Pay Gap Differences in asset composition and labor decisions in the two types of industries will affect gender pay gap if there is differential hiring decisions by gender in either industry, as we explained in SectionIV.1. Toexplorewhetherthereisdifferentialhiringpracticesbygender,wefirstexaminethe difference in labor share between the high-pay-gap and low-pay-gap industries for women and men before and after banking deregulation, as illustrated by the dotted red and blue lines, respectively, in Figure 4. The data shows that there was a sharp transition from the high-pay-gap to the lowpay-gap industries for women in the years after deregulation. While some men also transitioned towards the low-pay-gap industries immediately after deregulation, the extent of the transition is more muted, and men are equally represented in both types of industries in the subsequent years. Thisevidencesuggestsgendereddifferencesinhiringdecisionsinthehigh-pay-gapandlow-pay-gap industries in response to a change in credit condition. Tofurtherconfirmdifferencesinhiringdecisionsbygenderbetweenthetwotypesofindustries, we estimate the effects of bank deregulation on the probability of transitioning from the low-paygap to the high-pay-gap industries, or vice versa. The results are shown in columns (1)–(2) and columns (5)–(6) in Table (8). We measure industry-to-industry transition using a dummy variable that takes the value 1 for individuals who moved from a low-pay-gap to a high-pay-gap industry and vice versa during the previous year, and 0 otherwise. A negative estimate means that workers aremorelikelytostayinthesameindustry, andapositiveestimatemeansthattheyaremorelikely to transition. The results show that high-pay-gap industries are more likely to retain workers than medium-pay-gap industries by 5 log points following deregulation on average, while low-pay-gap industries are more likely to lose workers by about 7 log points. However, this pattern reverses when we zoom in on women. Relative to men, women are more likely to remain in low-pay-gap industries by 6 log points and more likely to leave high-pay-gap industries by 4 log points. In the presence of differential hiring patterns between men and women in the high-pay-gap industries, the low-pay-gap industries should compete more for female workers than male workers, whichwouldexertanupwardpressureontherelativewageforwomeninthelow-pay-gapindustries. As we discussed in Section III and shown in Table (4), relative wages for women increased by about 5%inthelow-pay-gapindustriesinresponsetobankingderegulation, whiletherewasnosignificant change in their pay in high-pay-gap industries. To further confirm that the change in relative wage for women is driven by differential hiring patterns in the two types of industries, we separately estimate Equation (1) for workers in 21

occupations with low or high risk of cross-industry transition. For occupations with high risk of cross-industry transition, banking deregulation should have a stronger effect on worker wage, as increasesinworkerdemandaremorelikelytospilloverfromindustrytoindustry. Wecategorizethe risk of transition for each occupation by the rate at which workers switch from low into high-paygap industries or vice versa. Occupations with switching rate less than the median are categorized as low-transition-risk occupations, and those with above median switching rate are categorized as high-transition-risk industries. The results are shown in Table (9). Columns (1)–(2) and (5)–(6) show the effects of intrastate and interstate deregulation, respectively, for workers in occupations with low risk of transitioning, and columns (3)–(4) and (7)–(8) show estimates for workers in occupations with high risk of transitioning. In the low-pay-gap industries, relative wages for women with low-transition-risk occupations risk increase by about 1–2% after deregulation but this increase is not statistically significant. However, for women with high-transition-risk occupations, relative wages significantly increase by 4–5%. The result that the relative increase in women’s wage is concentrated in high-transition-risk occupations support the idea that differential hiring patterns across industries is driving result. Furthermore, we evaluate how much compensation it takes to lure workers from the high-paygap to the low-pay-gap industries. As shown in columns (4)–(6) of Table (8), it takes an additional 5–6% increase in wages to lure a male worker from a high-pay-gap to a low-pay-gap following deregulation. For women, it takes only about 1–2%. In total, luring a male worker from a highpay-gap to a low-pay-gap industry would require an increase of 10–12%, while for women it takes only 1–3%. To summarize, we have shown that asset tangibility drive the effects of banking deregulation on gender inequality by (i) affecting firm borrowing differentially across industries, (ii) leading to differential changes in investment decisions and demand for workers, and (iii) creating imbalances in demand for workers across genders, which results in differential effects on gender pay gap in high-asset-tangibility and low-asset-tangibility industries. IV.5 Robustness In this subsection, we conduct three main sets of robustness analyses. We evaluate (i) potential alternativemechanismsdrivingourmainresultsontheeffectsofbankingderegulationongenderpay gap across industries; (ii) results based on alternative ways of categorizing industries, including by asset tangibility; and (iii) additional robustness tests controlling for industry-level characteristics. 22

Alternative Mechanisms. Supply-side channels may contribute to our main results on the differential effects of bank deregulation on gender pay gap in the low-pay-gap and high-pay-gap industries,aswellastheresultsondifferentialindustrytransitionsbetweenwomenandmen. Asone way to address this concern, we account for composition changes in the labor force by controlling for Mincerian traits × gender (education-gender, experience-gender, and experience2-gender) in all our specifications, following the standard practice in the literature. To further examine whether supply-side channels are playing a role, we perform a series of analyses. First,weexaminewhetherchangesinwagesareconsistentwithachangeindemandforworkers or a change in worker supply. For example, if there is a supply-side shift in preference toward a particular industry by one gender, then the relative wages for workers of that gender should decline; in contrast, if there is an increase in the demand for workers by a particular industry, wages in that industry should increase. Our results from Table (4) and Table (8) show that relative wages for women increases in the low-pay-gap industries while more women transition towards these industries. Taken together, these results suggest that it is unlikely that differential industry transitions and changes in wages between women and men are driven by supply-side forces. Second, we analyze whether credit expansion from bank deregulation differentially affected labor participation between women and men, which could affect their relative wages. As bank deregulation may change household lending, are banks lending disproportionately more to borrowers of a particular gender and thus generating differences in the labor participation patterns between these two groups? We test this conjecture by examining whether bank deregulation differentially affected labor market participation of a particular gender group by improving its (i) housing outcomes (residential choices allow moving into opportunity), (ii) transportation outcomes (easier commute allows better job prospects), and (iii) self-employment opportunities. In Appendix Tables (G.5) and (G.6), we evaluate the differential effects of intrastate and interstate deregulation, respectively, on housing and transportation outcomes using the CPS and Censusdata. Incolumns(1),(2),and(3),weevaluatetheeffectofderegulationonhomeownership, likelihood of moving into a different residence, and likelihood of holding a mortgage, respectively. Panels A, B, and C report the results for workers in all industries, the low-pay-gap industries, and the high-pay-gap industries, respectively. The coefficient of interest is Deregulation × Female. For all three housing outcome measures across all three panels, estimates are economically small and statistically indistinguishable from zero, which show that residential choices of female workers are not differentially affected by credit expansion from bank deregulation. In columns (4)–(5), we 23

conduct a similar analysis focusing on car ownership and transportation time to work (in minutes) as measures of work commute. Across all three panels, estimates of the coefficient of interest are economicallysmallandstatisticallyinsignificant, whichindicatethattransportationoutcomeswere not affected in a gendered way by deregulation. These two sets of results suggest that it is unlikely that differential access to credit between men and women is driving our main results. In Appendix Table (G.7), we show results on the effects of deregulation on self-employment incorporated rates (columns 1-3), self-employment unincoporated rates (columns 4-6), and incorporation rates conditional on self-employment (columns 7-9). Panel A reports the estimates from intrastate deregulation, and Panel B shows those from interstate deregulation. The coefficient of interest is again Deregulation×Female. In Panel A, we find that the effects of intrastate bank deregulationonself-employmentmeasuresbygenderarenotstatisticallysignificantoreconomically meaningful for any of the measures of self-employment. However, in Panel B we see that the effects of interstate bank deregulation are statistically significant and larger for workers in low-pay-gap industries (around 1% increase). Nevertheless, we do not think that the effects of interstate deregulation on self-employed incorporated rates by gender contribute to our main results in Table (4) for two reasons. First, the estimates in Table (4) are nearly identical for intrastate and interstate deregulation. If differential self-employment incorporated were a first-order driver of the main results, the effects on self-employment incorporated intrastate and interstate deregulation should be similar, but they are not. Moreover, the effects of deregulation on self-employment incorporated are close to zero for intrastate deregulation. Second, the difference in the estimates of interstate bank deregulation on self-employment incorporated by gender between the low-pay-gap and highpay-gapindustriesaresmallinmagnitude. Ifdifferentialself-employmentincorporatedwereamain driver of the main results, it must be the case that self-employment incorporated affects the main results differently in low-pay-gap and high-pay-gap industries. Alimitationofouranalysisisthatwecannotobservewhethertherelaxationofcreditconstraint helped individuals to invest in their skills in a gendered way, which could then affect differences in industrial-occupational choice across genders in a way that does not require divergent industrial responses. Nevertheless, there is indirect evidence challenging this conjecture. The initial changes in hiring patterns and relative wages in low-pay-gap industries were sharp (Figures 4 and 3). This is inconsistent with the conjecture of finance propelling gendered-differences in skill investments as a main explanation for our results, as investments in skills tend to occur with a time lag. This, of course, does not preclude the possibility that finance-propelled gendered investments in skills is a 24

complementary mechanism to the main mechanism put forward in this paper. Alternative Categorizations In section IV.1, we showed that there is a close relationship between gender pay gap and industries’ asset tangibility. Because this study aims to study the transformation of gender inequities, we have decided to, conceptually, focus on divergent industrial responses to deregulation along their preexisting gender pay gap levels. Nevertheless, we expect our results to be robust to categorizing industries by preexisting asset tangibility. To that end, we categorize industries into low asset tangibility and high asset tangibility based on the difference in the mean asset tangibility share in each industry during the pre-period of 1976–1980. The highasset-tangibility industries refer to industries that belong to the top 25% of the asset tangibility distribution, and the low-asset-tangibility industries refer to those in the bottom 25% of distribution. In the following, we show that our main results hold if we categorize industries by asset tangibility. Appendix Table (J.11) shows estimates from Equation (1) when industries are categorized by preexisting asset tangibility. We document that, following deregulation, wages increased in lowasset-tangibility industries (analogous to high-pay-gap industries), while wages in the high-assettangibilityindustries(analogoustolow-pay-gapindustries)andoverallwagesdeclined. Ouranalysis controls for county and year fixed effects as well as for Mincerian traits. Following deregulation, wages for workers in the low-asset-tangibility-industries increased by around 5–7% relative to other industries. This increase in wage is of similar magnitude to the estimates documented using our preferred categorization by preexisting pay gap levels. Changes in relative wages for women using the asset tangibility categorization also yields estimates similar to the ones documented in the main results. In the high-asset-tangibility industries (analogous to low-pay-gap industries), relative wagesforwomenincreasedbyaround3–5%inresponsetoderegulation. InAppendixTable(K.12), we show that the results on the effects of bank deregulation on firm borrowing are also robust to categorizing industries by low and high levels of asset tangibility. Other Industry-Level Robustness It is possible that fixed industrial characteristics differentially affect men and women in a way that is not triggered by deregulation. In particular, we aim to test whether the riskiness of an industry (proxied by earnings volatility or leverage) or the availability of growth opportunities (proxied by Tobins’ q) might explain the relative changes in wages for men and women. In Appendix Table (I.9), we show that the inclusion of these industry level characteristics (duly interacted with a female dummy indicator) does not significantly affect 25

our main results. V. Downstream Effects: Shaping Gender Norms We have shown that credit expansions, through gendered labor market dynamics, lead to gender differences in pay and in sorting across industries. In this section, we test whether these differences change views on gender norms. Papers have pointed out that gender norms may lower women’s wages and their labor market participation (Charles, Guryan, and Pan 2018) and affect women’s career choices (Crawford and MacLeod 1990; Ceci, Williams, and Barnett 2009; Bottia et al. 2015). Conversely, differences in sorting and opportunity cost, real and perceived, could create ripe conditions for the creation and reinforcement of gender norms. Workers, spouses, and observers may interpret the gender differences in pay and in sorting we document through gendered lens and assume biased views, or validate previously formed ones, on women and their role in the workplace. For example, they may regard women as less suitable for some jobs, as having a comparative advantage for staying at home, or as those whose careers should be subordinated to their husband’s. We test for such changes in views using data from the GSS. V.1 Empirical Specification and Variable Measures Specifically,weconjecturethattheeffectsofcreditexpansionongendernormsaremorepronounced in places with a bimodal industrial structure, in which there is a higher concentration of both lowpay-gap and high-pay gap industries, rather than nonbimodal industrial structures (e.g., industrial structures with only one type of industry). The gendered dynamics we document should be more pronouncedinabimodalindustrialstructurebecauseitallowsmoreopportunitiestoswitchbetween low-pay-gap and high-pay gap industries. To test our hypothesis on the effects of credit expansion on gender norms about the workplace, we estimate the following specification using the GSS: Sexism = α+β Spread ×DP +β DP +δ +γ +ε (2) irt 1 r rt 2 rt r t irt where Sexism is a measure of workplace sexism, Spread is a measure of the spread (or the degree of polarization) of available industrial choices for a worker, DP is a measure of credit expansion (or changes in bank deregulation) adapted for the geographic design of the GSS, and δ and γ denote r t year and region fixed effects, respectively. The coefficient of interest is β . 1 26

Measure of Workplace Sexism. We adopt a measure of workplace sexism following Charles, Guryan,andPan(2018). TheGSSasksitsrespondentsabouttheirattitudesonwomen’sroleinthe workplace, family, and society. We focus on responses to the three questions pertaining to beliefs about the role of women in the workplace: “Should women work?”; “Wife should help husbands career first.”; “Better for man to work, women tend home.” Respondents either approve/agree or disapprove/disagree with a given statement. For each question, we assign a value of one when the response reflects biased views against women and zero otherwise. To generate a standardized measure of sexism in the workplace, we then subtract individual responses to each question by the average response of entire population in 1977, a pre-treatment period, and divide them by the standard deviation of the initial response of the entire population in 1977, following Charles, Guryan, and Pan (2018). The standardized measure reflects where each individual belief stands in the spectrum of workplace sexism relative to the pre-treatment average. Measure of Industrial Spread. Wehypothesizethatchangesincreditconditionsaffectsgender norms through the gendered labor market dynamics we document and the resulting gendered sorting across the high- and low-pay-gap industries. Through this mechanism, the public’s views on gender roles should be affected more acutely in areas where gendered industrial composition is more pronounced and sorting is most likely to occur. When industrial composition in an area is characterized by a fifty-fifty split between jobs in the low-pay-gap and high-pay-gap industries, the differential opportunity cost of choosing an industry over another between men and women is at its highest. By comparison, when areas are dominated by a single type of industry, the differential opportunity cost must, trivially, be zero, as there is no de facto choice to be made. In short, higher industrial spread accentuates the dynamics of sorting, and lower industrial spread mitigates them. We proceed to formalize this notion in a measure that quantifies the degree of industrial spread within a geographic area. To measure the spread of industries in terms of pay gap, we classify each industry by the distance of its pay gap to the median-pay-gap industry. If an industry belongs to the top 25th percentile in terms of pay gap, i.e., the high-pay-gap industries, it is assigned a value of 1. If an industry belongs to the bottom 25th percentile, it is assigned a value of −1. Industries between the 25th and 75th percentiles, the median-pay-gap industries, are assigned a value of 0. Because the discrete value assigned to each industry represents its distance to the median-pay-gap industry, we can express the spread between industries as a composite of distances between any two industries. 27

For any two industries, the longest possible distance is 2. The spread is the expected value over pairwise combinations of workers. By taking the expected value, the largest possible spread is normalized to be 1. Formally, for every worker in a region, the overall industrial spread is the average pairwise distance between the industries of every two workers in a given region g: N 1 (cid:88) Spread = ∗ |x −x |, (3) g N2 i j ∀i,j∈g where x ∈ {−1,0,1} is the value of the industry in which worker i belongs, and N is the number i of workers in region g. As the spread increases, the margin for gendered dynamics to occur increases, which would lead to an environment more susceptible to the creation and reinforcement of gender norms. Measure of Deregulation Penetration. The GSS public data reports geographic affiliation of interviewee only at the region level. It divides the United States into nine different regions. Since bank deregulation changes occur at the state-level, we construct a penetration measure for each region-year to capture the proportion of the population affected by the new regulatory framework. This is, penetration refers to the proportion of individuals in region r affected by bank deregulation for each year t. Deregulation Penetration (DP) is defined as follows: (cid:88) pop st DP = D ∗ (4) rt st pop rt s∈r where pop denotes the population count living in state s in year t, pop denotes the total popst rt ulation living in region r in year t, and, as before, D is a dummy variable indicating whether st deregulation has taken place in year t. We use this measure as our treatment variable for credit expansion. V.2 Effects of Deregulation on Gender Norms We report the results based on Equation (2) in Table (10). We find that, following credit expansion, gender bias increases in areas with a higher degree of industrial spread between the high- and low-pay-gap industries, and this increase is driven mostly by men and households with children. In column (1), we find that following deregulation, workplace sexism in areas with industrial spread of 1, or a fully polarized geographical area, increased by 2.71 standard deviations relative to an area with an industrial spread of 0, or no polarization, based on our index of workplace sexism. 28

For households with children, workplace sexism increased by 3.27 standard deviations for areas with industrial spread of 1 (column 2). Both estimates are large and statistically significant. For reference, the average industrial spread in our sample is 0.75. One explanation for the stronger effects among people with children involves differential opportunity costs. As we previously document, the differences in earnings between the high-pay-gap and low-pay-gap industries are larger for men than for women and increase following deregulation. This means that the opportunity cost of staying at home also increases for men in places with the highest industrial spread, making households with children more likely to support gendered views about the workplace. We also run our analysis separately for men and women in columns (3)–(6) and (7)–(10), respectively. In particular, we focus on the responses to individual questions on workplace sexism in the survey in columns (4)–(6) and (8)–(10). We find that responses by men are driving the main overall effect. Following credit expansion, men are more likely to hold the views that women should not work, should prioritize their husband’s career, or should stay at home. The coefficients of interest across the three questions on workplace sexism are all large, statistically significant, and similar in magnitude. For women, the results across the three questions are more varied, revealing more complex views about the role of women in the workplace. Based on the results on theoverallindexofworkplacesexism,wefindthatwomen’sviewsongendernormsdidnotexhibita statisticallysignificantchangefollowingderegulation(albeitthecoefficientisstillpositive). Overall, the results indicate that gender norms about women in the workplace are mostly driven by males, and such views accentuated among male workers following the credit expansion. VI. Conclusion This paper proposes an asset channel of inequality that drives the persistence of gender inequities. We show that, through this channel, financial deregulation reduced the gender pay gap at the bottom of the pay gap distribution and induced gender sorting out of the top of the distribution. Specifically, we document that industries with high gender pay gaps have a low share of tangible assets, and industries with low gender pay gaps have a high share of tangible assets. Because asset tangibility determines firms’ collateral and ability to borrow, project selection, and labor demand, financial deregulation (which increases credit access) has different effects on workers who belong to industries with different levels of asset tangibility. In more equitable industries (i.e., industries with a lower pay gap and more tangible assets), firms increase borrowing and increases their demand for labor in response to financial deregulation. In more inequitable industries (i.e., 29

industries with a higher pay gap and more intangible assets), firms do not significantly change their borrowing but lower their demand for labor. Differences in labor demand between equitable industries and inequitable industries, together with higher relative pay for women in the more equitable industries, lead to gender sorting between the more equitable and inequitable industries. We further demonstrate that this sorting cements gender roles, which then accentuates workplace gender bias and reinforces glass ceilings. Our results have implications for understanding the evolution of the gender pay gap. Our findings suggest that the waves of financial deregulation in the 1980s contributed to the bottom-up narrowing of the gender pay gap by propelling a reduction in pay gap in lower-paying industries. In addition, these findings shed light on why gender inequities remain persistent by showing how relative gains that are heterogeneous across economy sectors can lead to gender sorting, and that this gender sorting across industries worsens sexism toward women. As gender roles cement, glass ceilings become harder to break. More broadly, the asset channel we document may play a role in other settings, affecting not only gender inequities but also other forms of inequality. Through this channel, credit conditions could trigger changes in labor market dynamics across industries, affecting workers in complex ways that could potentially compound preexisting inequities along different dimensions. These dimensions are potentially policy relevant and interesting avenues for further research. 30

References Aghion, Philippe, and Peter Howitt. 1992. “A Model of Growth Through Creative Destruction.” Econometrica: 323–351. Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith, and Peter Howitt. 2005a. “Competitionandinnovation:Aninverted-Urelationship.”The Quarterly Journal of Economics 120 (2): 701–728. Aghion, Philippe, Philippe Askenazy, Nicolas Berman, Gilbert Cette, and Laurent Eymard. 2012. “Credit constraints and the cyclicality of R&D investment: Evidence from France.” Journal of the European Economic Association 10 (5): 1001–1024. Aghion,Philippe,RobinBurgess,StephenRedding,andFabrizioZilibotti.2005b.“Entryliberalization and inequality in industrial performance.” Journal of the European Economic Association 3 (2-3): 291–302. Aghion, Philippe, Antonin Bergeaud, Gilbert Cette, R´emy Lecat, and H´el`ene Maghin. 2018. The inverted-U relationship between credit access and productivity growth.Technicalreport.London School of Economics and Political Science, LSE Library. Albanesi, Stefania. 2019. Changing Business Cycles: The Role of Women’s Employment. Working Paper, Working Paper Series 25655. National Bureau of Economic Research. Albanesi, Stefania, and Claudia Olivetti. 2009. “Home production, market production and the gender wage gap: Incentives and expectations.” Review of Economic dynamics 12 (1): 80–107. Almeida, Heitor, and Murillo Campello. 2007. “Financial constraints, asset tangibility, and corporate investment.” The Review of Financial Studies 20 (5): 1429–1460. Amel,DeanF.1993.Statelawsaffectingthegeographicexpansionofcommercialbanks.Unpublished Manuscript. Amore,MarioDaniele,CedricSchneider,andAlminasZˇaldokas.2013.“Creditsupplyandcorporate innovation.” Journal of Financial Economics 109 (3): 835–855. Ashenfelter, Orley, and Timothy Hannan. 1986. “Sex discrimination and product market competition: The case of the banking industry.” The Quarterly Journal of Economics 101 (1): 149– 173. Autor, David H, Frank Levy, and Richard J Murnane. 2003. “The skill content of recent technological change: An empirical exploration.” The Quarterly journal of economics 118 (4): 1279– 1333. Babcock, Linda, and Sara Laschever. 2003. Women don’t ask: Negotiation and the gender divide. Princeton University Press. Ballot, G´erard, Fathi Fakhfakh, and Erol Taymaz. 2006. “Who benefits from training and R&D, the firm or the workers?” British journal of industrial relations 44 (3): 473–495. Barth, Erling, Sari Pekkala Kerr, and Claudia Olivetti. 2017. The Dynamics of Gender Earnings Differentials: Evidence from Establishment Data.WorkingPaper,WorkingPaperSeries23381. National Bureau of Economic Research. Beck, Thorsten, Ross Levine, and Alexey Levkov. 2010. “Big Bad Banks? The Winners and Losers from Bank Deregulation in the United States.” The Journal of Finance 65 (5): 1637–1667. 31

Becker, Gary S. 1957. The Economics of Discrimination. Univ. Chicago Press. Bellucci,Andrea,AlexanderBorisov,andAlbertoZazzaro.2010.“Doesgendermatterinbank–firm relationships? Evidence from small business lending.” Journal of Banking & Finance 34 (12): 2968–2984. Bennedsen, Morten, Elena Simintzi, Margarita Tsoutsoura, and Daniel Wolfenzon. 2019. Do firms respondtogenderpaygaptransparency? WorkingPaper,WorkingPaperSeries25435.National Bureau of Economic Research. Bertrand, Marianne, Antoinette Schoar, and David Thesmar. 2007. “Banking Deregulation and Industry Structure: Evidence from the French Banking Reforms of 1985.” The Journal of Finance 62 (2): 597–628. Bewley, Truman F. 1995. “A depressed labor market as explained by participants.” The American Economic Review 85 (2): 250–254. Black, Sandra E, and Alexandra Spitz-Oener. 2010. “Explaining Women’s Success: Technological Change and the Skill Content of Women’s Work.” The Review of Economics and Statistics 92 (1): 187–194. Black, Sandra E., and Philip E. Strahan. 2002. “Entrepreneurship and Bank Credit Availability.” The Journal of Finance 57 (6): 2807–2833. . 2001. “The Division of Spoils: Rent-Sharing and Discrimination in a Regulated Industry.” American Economic Review 91 (4): 814–831. Blanchflower, David G, Phillip B. Levine, and David J. Zimmerman. 2003. “Discrimination in the small-business credit market.” Review of Economics and Statistics 85 (4): 930–943. Blau, Francine D., and Lawrence M. Kahn. 1997. “Swimming upstream: Trends in the gender wage differential in the 1980s.” Journal of labor Economics 15:1–42. . 2017. “The gender wage gap: Extent, trends, and explanations.” Journal of Economic Literature 55 (3): 789–865. Breza, Emily, Supreet Kaur, and Yogita Shamdasani. 2017. “The morale effects of pay inequality.” The Quarterly Journal of Economics 133 (2): 611–663. Card, David, Ana Rute Cardoso, and Patrick Kline. 2016. “Bargaining, sorting, and the gender wage gap: Quantifying the impact of firms on the relative pay of women.” The Quarterly Journal of Economics 131 (2): 633–686. Card, David, Francesco Devicienti, and Agata Maida. 2013. “Rent-sharing, holdup, and wages: Evidence from matched panel data.” Review of Economic Studies 81 (1): 84–111. Card, David, Alexandre Mas, Enrico Moretti, and Emmanuel Saez. 2012. “Inequality at work: The effect of peer salaries on job satisfaction.” American Economic Review 102 (6): 2981–3003. Cengiz, Doruk, Arindrajit Dube, Attila Lindner, and Ben Zipperer. 2019. “The effect of minimum wages on low-wage jobs.” The Quarterly Journal of Economics 134 (3): 1405–1454. Charles,KerwinKofi,JonathanGuryan,andJessicaPan.2018.The Effects of Sexism on American Women: The Role of Norms vs. Discrimination. Working Paper, Working Paper Series 24904. National Bureau of Economic Research. 32

Chava, Sudheer, Alexander Oettl, Ajay Subramanian, and Krishnamurthy V. Subramanian. 2013. “Banking deregulation and innovation.” Journal of Financial Economics 109 (3): 759–774. Clark, Andrew E, and Andrew J Oswald. 1996. “Satisfaction and comparison income.” Journal of public economics 61 (3): 359–381. Demirgu¨c-Kunt, Asli, and Vojislav Maksimovic. 1998. “Law, Finance, and Firm Growth.” The Journal of Finance 53 (6): 2107–2137. Demsetz, Harold. 1982. “Barriers to entry.” The American economic review 72 (1): 47–57. Goldin, Claudia. 2014. “A Grand Gender Convergence: Its Last Chapter.” American Ecnomic Review 104 (4): 1091–1119. . 2006. “The Quiet Revolution That Transformed Women’s Employment, Education, and Family.” American Ecnomic Review 96 (2): 1–21. Goldin, Claudia, Sari P. Kerr, Claudia Olivetti, and Erling Barth. 2017. “The Quiet Revolution ThatTransformedWomen’sEmployment,Education,andFamily.”American Ecnomic Review 107 (5): 110–114. Hall, Bronwyn H, and Josh Lerner. 2010. “The financing of R&D and innovation.” In Handbook of the Economics of Innovation, 1:609–639. Elsevier. Hart,Oliver,andJohnMoore.1994.“Atheoryofdebtbasedontheinalienabilityofhumancapital.” The Quarterly Journal of Economics 109 (4): 841–879. Jayaratne, Jith, and Philip E. Strahan. 1996. “The Finance-Growth Nexus: Evidence from Bank Branch Deregulation.” The Quarterly Journal of Economics 111 (3): 639–670. Jiang,Tianjiao,RossLevine,ChenLin,andLaiWei.2020.“Bankderegulationandcorporaterisk.” Journal of Corporate Finance 60:101520. King, Robert G., and Ross Levine. 1993. “Finance and Growth: Schumpeter Might Be Right.” The Quarterly Journal of Economics 108 (3): 717–737. Kline, Pat, Neviana Petkova, Heidi Williams, and Owen Zidar. Forthcoming. “Who Profits from Patents? Rent-sharing at Innovative Firms.” Quarterly Journal of Economics. Kroszner, Randall S, and Philip E Strahan. 1999. “What drives deregulation? Economics and politics of the relaxation of bank branching restrictions.” The Quarterly Journal of Economics 114 (4): 1437–1467. Lemmon, Michael, and Michael R. Roberts. 2010. “The Response of Corporate Financing and Investment to Changes in the Supply of Credit.” The Journal of Financial and Quantitative Analysis 45 (3): 555–587. Lev, Baruch. 2000. Intangibles: Management, measurement, and reporting. Brookings institution press. Levanon, Asaf, Paula England, and Paul Allison. 2009. “Occupational Feminization and Pay: Assessing Causal Dynamics Using 1950-2000 U.S. Census Data.” Social Forces 88 (2): 865–892. Mellor, Earl F, and Steven E Haugen. 1986. “Hourly paid workers: who they are and what they earn.” Monthly Lab. Rev. 109:20. 33

Nguyen,Hoai-LuuQ.2019.“AreCreditMarketsStillLocal?EvidencefromBankBranchClosings.” American Economic Journal: Applied Economics 11 (1): 1–32. Rajan, Raghuram, and Luigi Zingales. 1998. “Financial Dependence and Growth.” American Economic Review 88 (3): 559–86. Rice, Tara, and Philip E. Strahan. 2010. “Does Credit Competition Affect Small-Firm Finance?” The Journal of Finance 65 (3): 861–889. Shleifer, Andrei, and Robert W Vishny. 1992. “Liquidation values and debt capacity: A market equilibrium approach.” The Journal of Finance 47 (4): 1343–1366. Smirlock, Michael, Thomas Gilligan, and William Marshall. 1984. “Tobin’s q and the Structure- Performance Relationship.” The American Economic Review 74 (5): 1051–1060. Sun, Liyang, and Sarah Abraham. 2020. “Estimating dynamic treatment effects in event studies with heterogeneous treatment effects.” Journal of Econometrics. Van Reenen, John. 1996. “The Creation and Capture of Rents: Wages and Innovation in a Panel of U.K. Companies.” The Quarterly Journal of Economics 111 (1): 195–226. Williamson,OliverE.1988.“Corporatefinanceandcorporategovernance.”The Journal of Finance 43 (3): 567–591. 34

Figure 1: Industry Assets by Pay Gap Panel A: Asset Tangibility per Employee Panel B: Total Plant and Equipment per Employee Panel C: Total Assets per Employee Notes: Thisfigureplotsthreemeasuresofassetsforthelow-pay-gapandhigh-pay-gapindustriesbetween1980and2014usingCompustat. PanelAshowstotalassettangibilityperemployee;PanelBshowstotalplantandequipmentperemployee;andPanelCshowstotalassets per employee. Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of thepaygapdistribution,andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. 35

Figure 2: Industry Wage by Pay Gap Panel A: Average Industry Wage for the Low and High-Pay-Gap Industries Panel B: Differences in Median Wage between High and Low- Pay-Gap Industries by Gender Notes: Panel A plots the average industry wage for the high and low-pay-gap industries. Panel B plots the difference in median log wagebetweenthehigh-pay-gapandthelow-pay-gapindustriesbygender. Thedifferenceinmedianlogwagebetweenthetwoindustriesis computedbysubtractingthemedianlogwageofeachgenderinthelow-pay-gapindustriesfromthatofthesamegenderinthehigh-pay-gap industries. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlogwagebetweenmaleand female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the paygapdistribution,andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. Datasource: CPS. 36

Figure 3: Effects of Bank Deregulation on Gender Pay Gap: Event Studies Panel A: RelativeWagesforWomenintheLow-Pay-GapIndustries Panel B:AbsoluteWagesintheHigh-Pay-GapIndustries Notes: This figure plots coefficients from an event study version of Equation (1) that uses (log) wage as the dependent variable and includes state fixed effects, year fixed effects, and Mincerian controls. Panel A shows the coefficients on the interaction of female × dummies for years since deregulation in the low-pay-gap industries. Panel B shows the coefficients on the dummies for years since deregulation for the high-pay-gap industries. Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, and the low-pay-gap industries refer to those in the bottom 25% ofthepaygapdistribution. 37

Figure 4: Changes in Labor Force Participation in Low and High-Pay-Gap Industries Notes:Thisfigureplotsthelikelihoodofworkinginthehigh-andlow-pay-gapindustries10yearsbeforeand10yearsafterintrastate bankingderegulation(deregulationcorrespondstot=0),forallworkers(blackline)andbygender(womeninredandmeninblue), using raw CPS data. Workers in the low-pay-gap industries are assigned a value of −1; workers in the high-pay-gap industries are assignedavalueof1;andworkersinallotherindustriesareassignedavalueof0. Thelikelihoodofworkinginaparticularindustryis calculatedastheaverageoftheindicatorsineachperiod. Valuesgreaterthan0meanhigherlikelihoodofworkinginthehigh-pay-gap industries, and values less than 0 mean higher likelihood of working in the low-pay-gap industries. Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, and the low-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. 38

Figure 5: Effects of Bank Deregulation on R&D and Firm Revenue: Event Studies Panel A: R&DSpendingintheHigh-Pay-GapIndustries Panel B:RevenuePerWorkerintheHigh-Pay-GapIndustries Notes:ThisfigureplotscoefficientsonthedummiesforyearssincederegulationfromaneventstudyversionofEquation(1)thatuses R&D spending (Panel A) and firm revenue (Panel B) as the dependent variables and includes state fixed effects, year fixed effects, andMinceriancontrols. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlogwage betweenmaleandfemaleemployeesineachindustryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustriesthatbelongto thetop25%ofthepaygapdistribution,andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. 39

Table 1: Summary Statistics Panel A: Summary Statistics for Individuals (CPS) All Low High Men Women Men Women Men Women Wage(hourly) $13.65 $10.65 $11.61 $10.62 $16.54 $11.11 ($1.97) ($1.97) ($1.98) ($1.96) ($2.54) ($2.04) Education(years) 13.1 13.3 12.6 13.4 14.0 13.4 (2.9) (2.6) (3.2) (2.5) (2.7) (2.6) –HSGrad&Equiv(%) 21.7 22.4 22.6 22.6 15.1 20.0 (41.3) (41.7) (41.8) (41.8) (35.8) (40.0) –College(%) 16.6 18.2 13.7 18.0 24.8 19.4 (37.2) (38.6) (34.4) (38.5) (43.2) (39.5) –Post-College(%) 4.5 5.0 4.2 4.6 7.0 4.4 (20.7) (21.8) (20.1) (21.0) (25.5) (20.5) Age 40.7 40.2 40.1 40.2 40.9 39.7 (10.3) (10.2) (10.4) (10.3) (10.2) (10.1) Experience 27.6 26.9 27.4 26.8 26.9 26.3 (10.8) (10.8) (11.0) (10.8) (10.6) (10.8) Participation(%) 65.1 34.9 58.1 41.9 61.7 38.3 Panel B: Summary Statistics for Public Firms All Low High RevenueperEmployee($) 242.4 418.3 224.8 (1,055.8) (1,666.7) (852.4) NetIncomeperEmployee($) -31.7 -14.0 -45.7 (873.8) (652.7) (925.5) NetIncome+Operating 195.7 278.3 194.3 ExpenseperEmployee($) (910.0) (1,344.4) (737.2) Employees 6.0 5.5 4.5 (20.2) (15.0) (17.9) TotalAssets($) 1,325.8 1,367.3 1,326.4 (10,427.3) (6,376.7) (12,935.6) Tobin’sQ 1.02 0.92 1.09 (0.45) (0.37) (0.49) BookLeverage 0.51 0.55 0.47 (0.68) (1.43) (0.30) Tangibility 0.29 0.55 0.20 (0.24) (0.26) (0.17) Firms 10,089 1,612 5,981 Notes: This table reports summary statistics for the main analysis sample using the Current Population Survey (CPS) (Panel A) and Compustat(PanelB)from1976–2014. TheCPSmainsampleisrestrictedtoworking-agefull-timefull-yearworkersintheprivatesector excludingFIREindustries. Hourlywagesarederivedfromannualwageincome,usualweeklyhoursworked,andnumberofweeksworked. Tobin’sQ,bookleverage,andtangibilityaredefinedasfollows: Tobin’sQistheratiooftotalassets+sharesoutstanding×shareprice −commonequitytototalassets; bookleverageistheratioofshort-termdebt+long-termdebttoshort-termdebt+long-termdebt+ stockholdersequity;tangibilityistheratioofProperty,Plant,andEquipmenttototalassets. Foradditionaldetails,seeSectionII.2. 40

Table 2: Reliance on External Financing by Industries All Low High Mean sd Mean sd Mean sd PanelA:All Debt-to-Asset–Secured 0.085 0.144 0.125 0.174 0.061 0.119 Debt-to-Asset–Notes 0.066 0.120 0.106 0.152 0.045 0.096 Debt-to-Asset–Long-term 0.163 0.192 0.236 0.211 0.123 0.171 Leverage 0.496 0.270 0.533 0.266 0.459 0.270 PanelB:Pre-Deregulation Debt-to-Asset–Secured 0.106 0.152 0.128 0.174 0.085 0.127 Debt-to-Asset–Notes 0.085 0.127 0.105 0.148 0.065 0.105 Debt-to-Asset–Long-term 0.179 0.179 0.206 0.201 0.147 0.155 Leverage 0.507 0.252 0.510 0.282 0.482 0.238 PanelC:Post-Deregulation Debt-to-Asset–Secured 0.082 0.143 0.124 0.174 0.059 0.118 Debt-to-Asset–Notes 0.064 0.119 0.106 0.153 0.043 0.095 Debt-to-Asset–Long-term 0.161 0.193 0.242 0.213 0.122 0.172 Leverage 0.495 0.272 0.538 0.263 0.457 0.273 Notes: Thistablereportssummarystatisticsofdebt-to-assetratiosandleveragebyindustryusingCompustatdata. PanelAreportsthe averageandstandarddeviationfortheentiresampleperiodfrom1976to2014;PanelBreportsthosefortheperiodbeforederegulation; PanelCreportsthosefortheperiodafterderegulation. Fordetails,seeSectionII.2. 41

Table 3: Industry Descriptions Panel A: Highest and Lowest Pay Gap Industries Top10Industries Bottom10Industries OfficesandClinicsofDentists AgriculturalProduction,Crops OfficesandClinicsofPhysicians GasolineServiceStations LegalServices GrainMillProducts DrugStores ReligiousOrganizations ComputerandDataProcessingServices NursingandPersonalCareFacilities Advertising SocialServices MiscellaneousFabricatedTextileProducts HouseholdApplianceStores ManagementandPublicRelationsServices BeverageIndustries MiscellaneousProfessionalandRelatedServices OilandGasExtraction Accounting,Auditing,andBookkeepingServices ResidentialCareFacilities,withoutnursing Panel B: Fastest and Slowest Growing Industries Top10Industries PayGapLevel Bottom10Industries PayGapLevel Computeranddataprocessingservices High Privatehouseholds Med Agriculturalchemicals Low Agriculturalproduction,crops Low Research,development,andtestingservices Med Apparelandaccessories,exceptknit High Managementandpublicrelationsservices High Varietystores High Drugs High Footwear Low Electriclightandpower High Retailflorists Med Engineering,architectural,andsurveyingservices High Knittingmills Med Computersandrelatedequipment High Beautyshops Low Petroleumrefining High Eatinganddrinkingplaces Low Electricandgas,andothercombinations Med Laundry,cleaning,andgarmentservices High Notes: PanelAliststhetop10andbottom10industriesintermsofpaygap. PanelBliststhetop10andbottom10industriesinterms of employment growth. Pay gap is the difference between the mean log wage of male and female employees by industry during the years beforeandafterbankderegulationusingCPS.Thesampleisrestrictedtoindustriesthathiredatleast100femaleand100maleemployees during the sample period, which encompasses 105 industries (out of 189 total industries) in the CPS 1990 industry classification codes. Fordetails,seeSectionII.2. 42

Table 4: Effects of Bank Deregulation on Gender Pay Gap IntrastateDeregulation InterstateDeregulation (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -0.02 -0.02 -0.02 -0.02 -0.02∗∗ -0.02∗∗ -0.02∗ -0.02∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–LowPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.04∗∗∗ 0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighPGIndustry 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗-0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–LowPGIndustry -0.00 -0.00 0.00 0.00 0.01∗ 0.01∗ 0.01∗∗ 0.01∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–HighPGIndustry 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.12∗∗∗ 0.12∗∗∗ 0.12∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.13∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.02∗ -0.02∗ -0.03∗∗ -0.02∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.18∗∗∗ -0.18∗∗∗ -0.18∗∗∗ -0.18∗∗∗-0.19∗∗∗ -0.19∗∗∗ -0.19∗∗∗ -0.19∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) HighPGIndustry 0.03∗∗∗ 0.03∗∗∗ 0.03∗∗∗ 0.02∗∗∗ 0.02∗∗ 0.02∗∗ 0.02∗∗∗ 0.02∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) N 815,627 815,627 815,627 815,627 815,627 815,627 815,627 815,627 State×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender No Yes Yes Yes No Yes Yes Yes MaritalStatus×Gender No No Yes No No No Yes No Race×Gender No No No Yes No No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of the effects of bank deregulation on gender pay gap from Equation (1). Columns (1)–(4) report the effects of intrastate deregulation as the treatment, and columns (5)–(8) report the effects of interstate deregulationasthetreatment. Deregulationisadummyvariablethattakesthevalueonefortheyearsafterderegulationand0otherwise. Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. High-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. HighPGisadummyvariable thattakesthevalueoneforhigh-pay-gapindustriesand0otherwise. LowPGisadummyvariablethattakesthevalueoneforlow-pay-gap industriesand0otherwise. AllspecificationscontrolforMinceriantraits×gender,state×gender,andyear×genderfixedeffects. Columns (2)–(4) and (6)–(8) additionally control for age×gender fixed effects. Errors are clustered at the state level and reported in parentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 43

Table 5: Effects of Deregulation on Firm Borrowing DebtGrowth LongTermDebtGrowth DebtRatio (1) (2) (3) (4) (5) (6) (7) (8) (9) Intrastate–HighPGIndustry -0.02 -0.02 -0.02 -0.03 -0.03 -0.02 -0.19∗∗∗ -0.17∗∗∗ -0.15∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.05) (0.05) (0.05) Intrastate–LowPGIndustry 0.06∗∗ 0.05∗ 0.05∗ 0.07∗∗∗ 0.05∗∗ 0.05∗∗ -0.01 0.03 0.01 (0.03) (0.03) (0.02) (0.02) (0.02) (0.02) (0.08) (0.08) (0.08) Intrastate 0.04∗∗ 0.03 0.13∗∗ (0.02) (0.02) (0.05) N 65,379 65,330 64,283 65,432 65,383 64,317 65,422 65,373 64,323 Interstate–HighPGIndustry -0.03∗∗ -0.02 -0.02 -0.04∗∗ -0.03 -0.02 -0.13∗∗∗ -0.11∗∗∗ -0.10∗∗∗ (0.01) (0.02) (0.02) (0.02) (0.02) (0.02) (0.04) (0.04) (0.04) Interstate–LowPGIndustry 0.01 0.02 0.01 0.01 0.00 -0.00 -0.02 0.00 -0.02 (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) (0.06) (0.06) (0.06) Interstate 0.05∗∗∗ 0.06∗∗ 0.11∗∗∗ (0.02) (0.03) (0.04) N 65,379 65,330 64,283 65,432 65,383 64,317 65,422 65,373 64,323 FirmFX Yes Yes Yes Yes Yes Yes Yes Yes Yes YearFX Yes Yes Yes Yes Yes Yes Yes Yes Yes StateFX Yes Yes Yes Yes Yes Yes Yes Yes Yes State×YearFX No Yes Yes No Yes Yes No Yes Yes FirmControls No No Yes No No Yes No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportstheestimatesoftheeffectsofbankderegulationonfirmdebt. Thedependentvariableisdebtgrowthincolumns (1)–(3),long-termdebtgrowthincolumns(4)–(6),anddebtratioincolumns(7)–(9). Intrastateisadummyvariablethattakesthevalue onefortheyearsafterintrastatederegulationand0otherwise. Interstateisadummyvariablethattakesthevalueonefortheyearsafter interstatederegulationand0otherwise. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemean log wage between male and female employees in each industry during 1976–1980. High-pay-gap industries refer to industries that belong tothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. High PG is a dummy variable that takes the value one for high-pay-gap industries and 0 otherwise. Low PG is a dummy variable that takesthevalueoneforlow-pay-gapindustriesand0otherwise. Allspecificationscontrolforfirms,state,andyearfixedeffects. Columns (2),(5),(8),and(11)additionallycontrolforstate×yearfixedeffects. Errorsareclusteredatthestatelevelandreportedinparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 44

Table 6: Effects of Deregulation on Tangible Asset Share by Firm %Tangible R&DSpending (1) (2) (3) (4) (5) (6) Intrastate–HighPGIndustry -0.00 -0.01 -0.01 0.31∗∗∗ 0.32∗∗∗ 0.32∗∗∗ (0.00) (0.01) (0.00) (0.06) (0.05) (0.03) Intrastate–LowPGIndustry 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗∗ -0.09 0.01 0.09 (0.01) (0.01) (0.01) (0.08) (0.09) (0.07) Intrastate -0.00 -0.29∗∗∗ (0.01) (0.06) N 68,407 68,355 60,593 41,535 41,387 36,541 Interstate–HighPGIndustry -0.01∗∗ -0.01∗∗ -0.01∗∗∗ 0.35∗∗∗ 0.33∗∗∗ 0.29∗∗∗ (0.00) (0.00) (0.00) (0.05) (0.04) (0.03) Interstate–LowPGIndustry 0.02∗∗∗ 0.02∗∗∗ 0.01∗∗∗ -0.12∗ -0.07 0.01 (0.00) (0.01) (0.01) (0.07) (0.08) (0.06) Interstate -0.00 -0.22∗∗∗ (0.00) (0.04) N 68,407 68,355 60,593 41,535 41,387 36,541 FirmFX Yes Yes Yes Yes Yes Yes YearFX Yes Yes Yes Yes Yes Yes StateFX Yes Yes Yes Yes Yes Yes State×YearFX No Yes Yes No Yes Yes FirmControls No No Yes No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the estimates of the effects of bank deregulation on asset tangibility and R&D. The dependent variable is the log(tangibleassets/totalassets)incolumns(1)–(3)andlog(R&Dexpenditure)incolumns(4)–(6)). Intrastateisadummyvariablethat takes the value one for the years after intrastate deregulation and 0 otherwise. Interstate is a dummy variable that takes the value one for the years after interstate deregulation and 0 otherwise. Industries are categorized into low pay gap and high pay gap based on the differenceinthemeanlogwagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. High-pay-gapindustriesreferto industriesthatbelongtothetop25%ofthepaygapdistribution,andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthe pay gap distribution. High PG is a dummy variable that takes the value one for high-pay-gap industries and 0 otherwise. Low PG is a dummy variable that takes the value one for low-pay-gap industries and 0 otherwise. All specifications control for firms, state, and year fixed effects. Columns (2), (5), (8), and (11) additionally control for state×year fixed effects. Errors are clustered at the state level and reportedinparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 45

eeyolpmE reP euneveR no noitalugereD fo stceffE :7 elbaT eeyolpmErePemocnIteN esnepxEgnitarepO+emocnIteN eeyolpmErePeuneveR tnemyolpmE eeyolpmEreP )21( )11( )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( ∗∗90.0 50.0 70.0 ∗∗∗31.0 ∗∗∗90.0 ∗∗∗90.0 ∗∗∗31.0 ∗∗90.0 ∗∗∗90.0 40.0- 40.0- 40.0yrtsudnIGPhgiH–etatsartnI )40.0( )50.0( )40.0( )30.0( )30.0( )30.0( )30.0( )40.0( )30.0( )30.0( )30.0( )30.0( 40.0 80.0- 11.0- 00.0 ∗∗∗31.0- ∗∗∗31.0- 00.0 ∗∗∗21.0- ∗∗∗21.0- ∗∗70.0 ∗∗80.0 ∗∗∗11.0 yrtsudnIGPwoL–etatsartnI )50.0( )60.0( )70.0( )20.0( )30.0( )30.0( )30.0( )30.0( )30.0( )40.0( )40.0( )40.0( 10.0- 30.0- ∗50.0- 40.0etatsartnI )40.0( )20.0( )20.0( )30.0( 265,64 963,25 354,25 685,16 715,96 675,96 515,26 096,07 847,07 075,96 248,07 998,07 N ∗70.0 40.0 ∗∗50.0 ∗∗∗11.0 ∗∗∗11.0 ∗∗∗21.0 ∗∗∗21.0 ∗∗∗21.0 ∗∗∗31.0 40.0- ∗∗50.0- ∗∗∗60.0yrtsudnIGPhgiH–etatsretnI )30.0( )30.0( )20.0( )30.0( )30.0( )20.0( )30.0( )30.0( )20.0( )20.0( )30.0( )20.0( 20.0 11.0- ∗∗51.0- 30.0- ∗∗41.0- ∗∗∗61.0- 10.0- ∗∗∗11.0- ∗∗∗31.0- ∗∗∗01.0 ∗∗∗11.0 ∗∗∗31.0 yrtsudnIGPwoL–etatsretnI )60.0( )70.0( )60.0( )40.0( )50.0( )50.0( )40.0( )40.0( )40.0( )30.0( )30.0( )30.0( 30.0 10.0- ∗40.0- 10.0 etatsretnI )30.0( )30.0( )20.0( )10.0( 265,64 963,25 354,25 685,16 715,96 675,96 515,26 096,07 847,07 075,96 248,07 998,07 N :XFlevel-mriF seY seY seY seY seY seY seY seY seY seY seY seY XFmriF seY seY seY seY seY seY seY seY seY seY seY seY XFraeY seY seY seY seY seY seY seY seY seY seY seY seY XFetatS seY seY oN seY seY oN seY seY oN seY seY oN XFraeY×etatS seY oN oN seY oN oN seY oN oN seY oN oN slortnoCmriF 10.0<p∗∗∗,50.0<p∗∗,01.0<p∗ rep emocni ten dna ,eeyolpme rep seunever ,tnemyolpme no noitalugered knab fo stceffe eht fo setamitse eht stroper elbat sihT :setoN fo rebmun / eunever(gol ,)3(–)1( snmuloc ni )seeyolpme fo rebmun(gol si elbairav tnedneped ehT .level etats ta yrtsudni yb eeyolpme emocni ten(gol dna ,)9(–)7( snmuloc ni )seeyolpme fo rebmun / ]esnepxe gnitarepo + emocni ten[(gol ,)6(–)4( snmuloc ni )seeyolpme etatsretni retfa sraey eht rof eno eulav eht sekat taht elbairav ymmud a si etatsretnI .)21(–)01( snmuloc ni )seeyolpme fo rebmun / egawgolnaemehtniecnereffidehtnodesabpagyaphgihdnapagyapwolotnidezirogetaceraseirtsudnI .esiwrehto0dnanoitalugered potehtotgnolebtahtseirtsudniotreferseirtsudnipag-yap-hgiH .0891–6791gnirudyrtsudnihcaeniseeyolpmeelamefdnaelamneewteb siGPhgiH .noitubirtsidpagyapehtfo%52mottobehtniesohtotreferseirtsudnipag-yap-wolehtdna,noitubirtsidpagyapehtfo%52 eulavehtsekattahtelbairavymmudasiGPwoL .esiwrehto0dnaseirtsudnipag-yap-hgihrofenoeulavehtsekattahtelbairavymmuda dna,)8(,)5(,)2(snmuloC .stceffedexfiraeydna,etats,smrfiroflortnocsnoitacfiicepsllA .esiwrehto0dnaseirtsudnipag-yap-wolrofeno *** dna ,**,* .sesehtnerap ni detroper dna level etats eht ta deretsulc era srorrE .stceffe dexfi raey×etats rof lortnoc yllanoitidda )11( .ylevitcepser,slevel%1dna,%5,%01ehttaecnacfiingisetacidni 46

snoitisnarT no noitalugereD fo stceffE :8 elbaT etatsretnI etatsartnI egaW∆ noitisnarTyrtsudnI egaW∆ noitisnarTyrtsudnI )8( )7( )6( )5( )4( )3( )2( )1( 30.0- 20.0- ∗∗∗50.0- ∗∗∗50.0- 00.0 00.0 ∗∗∗60.0- ∗∗∗60.0yrtsudnIGPwoL–elameF×noitalugereD )50.0( )50.0( )10.0( )10.0( )40.0( )40.0( )10.0( )10.0( 80.0- 70.0- ∗∗∗40.0 ∗∗∗40.0 40.0- 40.0- ∗∗∗40.0 ∗∗∗40.0 yrtsudnIGPhgiH–elameF×noitalugereD )50.0( )50.0( )00.0( )00.0( )40.0( )30.0( )00.0( )00.0( 20.0- 20.0- ∗∗∗60.0 ∗∗∗60.0 20.0- 20.0- ∗∗∗70.0 ∗∗∗70.0 yrtsudnIGPwoL–noitalugereD )20.0( )20.0( )10.0( )10.0( )20.0( )20.0( )10.0( )10.0( 30.0 30.0 ∗∗∗40.0- ∗∗∗40.0- ∗30.0 ∗40.0 ∗∗∗50.0- ∗∗∗50.0yrtsudnIGPhgiH–noitalugereD )20.0( )20.0( )00.0( )00.0( )20.0( )20.0( )10.0( )10.0( ∗∗∗50.0- ∗∗∗50.0- ∗∗∗80.0- ∗∗∗80.0- ∗∗50.0- ∗∗50.0- ∗∗∗90.0- ∗∗∗90.0yrtsudnIGPhgiH–elameF )20.0( )20.0( )10.0( )10.0( )20.0( )20.0( )10.0( )10.0( ∗∗∗50.0 ∗∗∗60.0 ∗∗∗21.0 ∗∗∗21.0 ∗∗∗50.0 ∗∗∗60.0 ∗∗∗41.0 ∗∗∗41.0 yrtsudnIGPhgiH )20.0( )20.0( )10.0( )10.0( )20.0( )20.0( )10.0( )10.0( 115,02 115,02 659,905 659,905 115,02 115,02 659,905 659,905 N seY seY seY seY seY seY seY seY redneG×ytnuoC seY seY seY seY seY seY seY seY redneG×raeY seY oN seY oN seY oN seY oN slortnoC 10.0<p ∗∗∗ ,50.0<p ∗∗ ,01.0<p ∗ emas eht gnisu ytilibom srekrow no noitalugered knab fo stceffe eht fo setamitse secnereffid-ni-ecnereffid eht stroper elbat sihT :setoN eulav eht sekat taht ymmud a si elbairav tnedneped eht ,)6(–)5( dna )2(–)1( snmuloc nI .)1( noitauqE ni srotacidni dna slortnoc fo tes eht ,)8(–)7( dna )4(–)3( snmuloc nI .asrev eciv ro seirtsudni pag-yap-hgih eht ot pag-yap-wol eht morf noitisnart ohw srekrow rof eno si noitalugereD .seirtsudni pag-yap-wol dna -hgih eht neewteb denoitisnart ohw esoht rof )egaw(gol ni egnahc eht si elbairav tnedneped pag yap wol otni dezirogetac era seirtsudnI .esiwrehto 0 dna noitalugered retfa sraey eht rof eno eulav eht sekat taht elbairav ymmud a .0891–6791 gnirud yrtsudni hcae ni seeyolpme elamef dna elam neewteb egaw gol naem eht ni ecnereffid eht no desab pag yap hgih dna otreferseirtsudnipag-yap-wolehtdna,noitubirtsidpagyapehtfo%52potehtotgnolebtahtseirtsudniotreferseirtsudnipag-yap-hgiH dnaseirtsudnipag-yap-hgihrofenoeulavehtsekattahtelbairavymmudasiGPhgiH .noitubirtsidpagyapehtfo%52mottobehtniesoht lortnocsnoitacfiicepsllA .esiwrehto0dnaseirtsudnipag-yap-wolrofenoeulavehtsekattahtelbairavymmudasiGPwoL .esiwrehto0 .slortnoc ecar dna egairram dda snmuloc derebmun-nevE .stceffe dexfi redneg×raey dna redneg×etats dna ,redneg×stiart nairecniM rof eht ta ecnacfiingis etacidni *** dna ,**,* .sesehtnerap ni detroper dna level etats eht ta deretsulc era srorrE .?? noitceS ees ,sliated roF .ylevitcepser,slevel%1dna,%5,%01 47

Table 9: Effects of Deregulation on Gender Pay Gap By Risk of Transition Intrastate Interstate LowRisk HighRisk LowRisk HighRisk (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -0.01 -0.01 -0.02 -0.02 -0.01 -0.01 -0.03 -0.02 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Deregulation×Female–LowPGIndustry 0.02 0.01 0.05∗∗∗ 0.04∗∗∗ 0.01 0.01 0.03∗∗ 0.03∗ (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighPGIndustry 0.01 0.01 0.02 0.01 0.02∗ 0.02∗ -0.00 0.00 (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Deregulation -0.03∗∗ -0.03∗∗ -0.04∗∗∗ -0.04∗∗∗ -0.06∗∗∗ -0.06∗∗∗ -0.05∗∗∗ -0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–LowPGIndustry 0.01 0.01 0.01 0.02 0.02∗ 0.02∗ 0.04∗∗∗ 0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–HighPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.09∗∗∗ 0.09∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.13∗∗∗ 0.13∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.16∗∗∗ 0.15∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.15∗∗∗ 0.14∗∗∗ (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.02∗ -0.03∗∗ -0.08∗∗∗ -0.08∗∗∗ -0.02∗∗ -0.02∗∗ -0.09∗∗∗ -0.10∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.11∗∗∗ -0.11∗∗∗ -0.25∗∗∗ -0.24∗∗∗ -0.12∗∗∗ -0.12∗∗∗ -0.26∗∗∗ -0.25∗∗∗ (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) HighPGIndustry 0.01 0.01 0.06∗∗∗ 0.06∗∗∗ 0.01 0.01 0.05∗∗∗ 0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) N 405,392 405,392 391,373 391,373 402,088 402,088 400,529 400,529 State×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender No Yes Yes Yes No Yes Yes Yes MaritalStatus×Gender No Yes Yes Yes No Yes Yes Yes Race×Gender No Yes Yes Yes No Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of Equation (1) for workers with occupations at low or high risk of cross-industrytransitions. Columns(1)–(3)reportstheestimatesforworkerswithoccupationsinthelow-transition-riskgroup. Columns (4)–(6)reportstheestimatesforworkerswithoccupationsinthehigh-transition-riskgroup. Deregulationisadummyvariablethattakes thevalueonefortheyearsafterderegulationand0otherwise. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthe differenceinthemeanlogwagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. High-pay-gapindustriesreferto industriesthatbelongtothetop25%ofthepaygapdistribution,andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthe pay gap distribution. High PG is a dummy variable that takes the value one for high-pay-gap industries and 0 otherwise. Low PG is a dummyvariablethattakesthevalueoneforlow-pay-gapindustriesand0otherwise. AllspecificationscontrolforMinceriantraits×gender, state×gender, and year×gender fixed effects. Columns (2) and (4) additionally control for age×gender, marital status×gender, and race×gender fixed effects. Errors are clustered at the state level and reported in parentheses. *,**, and *** indicate significance at the 10%,5%,and1%levels,respectively. 48

smroN redneG no noitalugereD fo stceffE :01 elbaT nemoW neM nerdlihChtiW llA nemoW dnabsuH dluohSnemoW ecalpkroW nemoW dnabsuH dluohSnemoW ecalpkroW ecalpkroW ecalpkroW emoHyatS tsriFreeraC kroWtoN msixeS emoHyatS tsriFreeraC kroWtoN msixeS msixeS msixeS )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( noitalugereDetatsartnI:AlenaP 27.2- 61.0- ∗∗∗52.3 25.1 ∗54.4 ∗∗87.4 ∗∗57.4 ∗∗∗63.4 ∗∗∗72.3 ∗∗∗17.2 daerpSlairtsudnI×etatsartnI )40.2( )50.2( )22.1( )31.1( )73.2( )38.1( )81.2( )36.1( )09.0( )19.0( ∗∗∗02.0- ∗11.0- ∗∗∗31.0- ∗∗∗21.0- ∗21.0- 01.0- 90.0- ∗∗21.0- ∗∗80.0- ∗∗∗21.0noitartenePetatsartnI )60.0( )70.0( )50.0( )40.0( )60.0( )70.0( )60.0( )50.0( )30.0( )30.0( 530,61 707,8 933,11 140,91 594,21 406,6 475,8 547,41 484,42 687,33 N noitalugereDetatsretnI:BlenaP 40.1 22.1 ∗82.2 ∗∗20.2 ∗∗01.6 ∗∗∗85.5 ∗91.3 ∗∗27.3 ∗∗∗29.2 ∗∗37.2 gnidaerpSlairtsudnI×etatsretnI )17.2( )86.2( )91.1( )89.0( )36.2( )18.1( )08.1( )16.1( )99.0( )90.1( 50.0- 31.0- ∗∗61.0- 70.0- 01.0- 70.0- ∗∗71.0- 50.0- 10.0- 60.0noitartenePetatsretnI )01.0( )80.0( )70.0( )50.0( )70.0( )90.0( )80.0( )50.0( )50.0( )40.0( 530,61 707,8 933,11 140,91 594,21 406,6 475,8 547,41 484,42 687,33 N seY seY seY seY seY seY seY seY seY seY XFraeY seY seY seY seY seY seY seY seY seY seY XFnoigeR 10.0<p∗∗∗,50.0<p∗∗,01.0<p∗ ehT .atad yevrus )SSG( yevruS laicoS lareneG eht gnisu level lanoiger ta smron redneg no noitalugered knab fo stceffe eht fo setamitse stroper elbat sihT :setoN eht ni nemow fo elor eht tuoba sfeileb ot gniniatrep snoitseuq eerht eht ot esnopser no desab msixes ecalpkrow fo serusaem dezidradnats era selbairav tnedneped fo eulav a ngissa ew ,noitseuq hcae roF ”.emoh dnet nemow ,krow ot nam rof retteB“ ;”.tsrfi reerac sdnabsuh pleh dluohs efiW“ ;”?krow nemow dluohS“ :ecalpkrow ybnoitseuqhcaeotsesnopserlaudividnignitcartbusybseulavehtezidradnatsnehteW .esiwrehtoorezdnanemowtsniagasweivdesaibstcefleresnopserehtnehweno ni noitalupop eritne eht fo esnopser laitini eht fo noitaived dradnats eht yb ti gnidivid dna ,doirep tnemtaert-erp a ,7791 ni noitalupop eritne fo esnopser egareva eht smret ni elitnecrep ht52 pot eht ot sgnoleb ti fI .yrtsudni pag-yap-naidem eht ot pag yap sti fo ecnatsid eht yb yrtsudni hcae yfissalc ew ,daerps lairtsudni roF .7791 eht neewteb seirtsudnI .1? fo eulav a dengissa si ti ,elitnecrep ht52 mottob eht ot sgnoleb ti fI .1 fo eulav a dengissa si ti ,seirtsudni pag-yap-hgih eht ,.e.i ,pag yap fo desab srekrow fo snoitanibmoc esiwriap revo eulav detcepxe eht si daerps lairtsudnI .0 fo eulav a nevig era ,seirtsudni pag-yap-naidem eht ,selitnecrep ht57 dna ht52 noitauqE no desab detaluclac ,noitalugered knab yb detceffa raey-noiger hcae ni slaudividni fo noitroporp eht fo erusaem a si noitartenep noitalugereD .3 noitauqE no srorre deppartstooB .stceffe dexfi noiger dna stceffe dexfi raey rof lortnoc snoitacfiiceps llA .noitcurtsnoc elbairav fo snoitpircsed deliated erom rof 1.V noitceS eeS .4 .ylevitcepser,slevel%1dna,%5,%01ehttaecnacfiingisetacidni***dna,**,* .sesehtnerapnidetroper 49

Appendices Table of Contents A DivergentIndustrialResponsestoCreditExpansions: ConceptualFramework 51 B Balance 52 B.1 Balance in Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 B.2 Balance in Covariates’ Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 C Occupational Differences Across Industries 52 C.1 Unequal Industries’ Worker Skills are Mostly Nonroutine Cognitive; Equitable Industries’ Worker Skills are Mostly Nonroutine Manual . . . . . . . 52 D Bottom-Up Convergence in Pay? Oaxaca-Blinder Decompositions 53 E Reversal of Fortune: Vulnerability to Credit Contractions 53 F Robustness of Industry Equitability Categorization 55 G Effects on Direct Lending to Worker 55 G.1 Effects of Bank Deregulation on Gender Differences in Housing and Transportation 55 G.2 Effects of Bank Deregulation on Gender Differences in Self-Employment . . . . . . 55 H Routine vs. Nonroutine Workers 56 I Controlling for Tobin’s Q, Earnings Volatility and Leverage by Gender 56 J Alternative Categorizations 56 J.1 AnalysiswithCategorizationbyIndustrialGenderPayGapin1968–1972(Instead of 1976–1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 J.2 Analysis with Categorization by Asset Tangibility . . . . . . . . . . . . . . . . . . 57 K More on Asset Tangibility 57 L Unstaggered Difference-in-differences Estimates 57 50

A. Divergent Industrial Responses to Credit Expansions: Conceptual Framework Weprovideaframeworktostudywhatoccurstolabormarketswhencreditexpansionsdifferentially opens up the availability of new positive NPV projects or ventures. Consider two industries i ∈ {L,H} and j. Every period, each industry engages in routine ventures and new ventures. New ventures require extra capital but generate monopolistic profits π per worker. Firms hire workers for both routine and new ventures. Workers in routine ventures i are paid at a competitive spot wage wT. Workers for new venture roles must be trained for a time t before working. A subset of workers of size K also are trainable in nonroutine tasks. Crossindustry competition for workers occurs only through the trainability dimension and not through other aptitudes. New ventures can only hire trainable workers, and try to hire M < K trainable 2 workers. The cost associated with training a worker, tπ , is akin to a replacement cost and generates i a quasi-rent that must be bargained between the worker and the firm in a bilateral monopoly.17 Bargaining is governed by a Nash protocol where workers have bargaining power β. Industry H is less competitive than industry L, and thus generates higher profits from the new ventures than industry L, i.e., π > π .18 The wage for a trainable worker in industry H is: H L wI = wT +βtπ H H This is, the worker is more productive in a new venture and captures part of that productivity in the form of higher wages, which is consistent with the findings of Van Reenen (1996) documenting a large rent-sharing elasticity in innovative firms. Group membership g ∈ {a,b} need not be correlated with the productivity of workers. Without loss of generality assume group membership is orthogonal to worker productivity,19 and that the owners of the means of production in industry H do not have any monetary incentive to preferentially hire from any group, but neither do they prevent preferential hiring by group. For idiosyncratic reasons, during the hiring process in industry H, workers of group a are weakly preferred to workers of group b, with a positive probability that group a is strictly prefer to group b.20 New ventures in industry H hire λM workers of group a and (1−λ)M workers of group b. NewventuresinindustryLpaywagesthatareabovethetraditionalmarketwage, butthatare strictly below wages at new ventures in industry H. Workers can be poached. As a consequence, they are not indifferent to hiring decisions in industry H. In particular, industry L sets wages such that they are indifferent between hiring a worker of group a or b. This implies that: (cid:16) M (cid:17) wI = wT +βtπ 1− (1−λ) L,b L K 17. ThereplacementcostassumptioninouranalysisisconsistentwithrecentempiricalevidencefromKlineetal., 2018, who show that firms disproportionately share rents with workers with high replacement costs. More on this below. 18. This is due to pledgeability differences explained in subsection IV.1. 19. Skills correlated with group membership amplify this problem. For example, highly educated workers may be deemed overqualified relative to workers with similar experience in a task with low complexity, especially when the replacement cost is high (Bewley 1995). 20. Wemakethisassumptionwithoutlossofgeneralitysincetheproblemissymmetricformembersofgroupaor b. 51

(cid:16) M (cid:17) wI = wT +βtπ 1− λ L,a L K and therefore: M(cid:16) (cid:17) wI −wI = βtπ 2λ−1 > 0. L,b L,a L K We summarize as follows: Pay Gaps In High and Low Surplus Industries. When an industry with high surplus disproportionately shares rent with workers of group a, industries with low surplus will find optimal to disproportionately hire or pay more to workers of group b. Thereplacementcostassumptionfindssupportinrecentempiricalevidence. Klineetal. (2018) find that firms disproportionately share rents with workers with high replacement costs, and that these workers are mostly men. Since the group disproportionately benefited is men, according to our framework, women will benefit in industries with lower surpluses. We will test this throughout the paper. Itisimportanttoremarkthatthecross-industrydynamicsnotonlyapplytowhengrouprefers to gender; it extends to multiple other dimensions documented to matter in the labor market. For example, our framework predicts a set of findings in Beck et al. (2010), in which that the value of other noneducation characteristics, e.g. experience, for low pay jobs should increase if demand for another proxy for skill, e.g. education, increases in high paid jobs. In that finding, group a refers to workers with high education, and group b are workers with low education (but other noneducation traits). Beck et al. (2010) overall finding is that following deregulation inequality decreases, converging from the bottom of the education distribution. The findings of this paper connect the findings of Beck et al. (2010) with those of Blau and Kahn (1997) by showing that deregulation generates bottom-up convergence in the gender pay gap. B. Balance B.1 Balance in Covariates B.2 Balance in Covariates’ Trends C. Occupational Differences Across Industries C.1 Unequal Industries’ Worker Skills are Mostly Nonroutine Cognitive; Equitable Industries’ Worker Skills are Mostly Nonroutine Manual In terms of employment composition, the high- and low-pay-gap industries differ mostly along their nonroutine skills. The high-pay-gap industries employ a labor force with high levels of nonroutine cognitive skills while the low-pay-gap industries employ mostly nonroutine manual skill workers (Figure 3). This is consistent with both the high levels of intangible assets and the low levels of external financing in the high-pay-gap industries (Hart and Moore 1994). In terms of routine skills, routinecognitiveandroutinemanualskillswerelargelyconcentratedinthehigh-pay-gapindustries at the start of the sample period, but over the ensuing decades, the share of routine workers in the these industries steadily declined, converging toward that of the low-pay-gap industries. The highand low-pay-gap industries differ mostly along their nonroutine skills. 52

Mellor and Haugen (1986) document that, in 1984, non-hourly paid workers work more hours than hourly paid workers, with women overrepresented in hourly paid positions. In our distinction between the high- and low-pay-gap industries, we document similar findings. Men are only slightly more likely to work in the high-pay-gap industries than women, but those differences accentuate when we focus on type of work. Women are disproportionately overrepresented in hourly paid work (Figure C.5). These differences are consistent with the findings of Goldin (2014), which emphasize the role of long working hours and temporal flexibility in explaining the gender pay gap. D. Bottom-Up Convergence in Pay? Oaxaca-Blinder Decompositions Our results demonstrate convergence in pay in the low-pay-gap industries. Does the convergence lead to an unambiguous reduction in the gender pay gap? To address this question, we perform two simple Oaxaca-Blinder decompositions, estimated one year before and five years after banking deregulation. The results are shown in Appendix Table (1). In columns (1), (3), and (5), we show that the low-pay-gap industries’ contribution to the overall pay gap lowers from −0.013 log points prederegulation to −0.039 log points post-deregulation, netting to a reduction of −0.026 log points. During the same period, high-pay-gap industries’ contribution to the pay gap lowers from +0.034 log points to +0.022 log points. Overall, the net effect is a reduction of 0.038 log points in the pay gap or about 34.4% of the total decline during that period. The effects are mostly bottom-up driven. Out of the 34.4% contribution, 69% is driven by the low-pay-gap industries. Moreover, after deregulation, the low-pay-gap industries explain −12.5% of the pay gap—that is a −9.4% change from pre-deregulation levels. In contrast, the high-pay-gap industries after deregulation still contribute +7.0% to the pay gap. Blau and Kahn (1997) find that the gender pay gap converges despite rising labor market inequality. In Table (D.1), we show that this finding also holds in our setting. E. Reversal of Fortune: Vulnerability to Credit Contractions We have shown that credit liberalization increase relative wage for women in the low-paying lowpay-gap industries. These increases do not stem from higher revenues in these industries but from the response of the low-pay-gap industries to higher revenues in the already high-paying industries. A natural ensuing question is whether these gains are permanent. More specifically, if an easing of credit access reduce the pay gap for women in some industries, do credit contractions have the opposite effect—are women’s wages more vulnerable to credit contractions? Additional Data Sources For our analysis on credit contractions, we use bank mergers that led to branch closings as our treatment. We use two alternative methods to pinpoint mergers that work as credit supply shocks. For both methods we restrict to mergers occurring during the 2000s but prior to the Great Recession, in order to avoid capturing many of the mergers that occurred because of the recession. We use the FDIC Call Reports and Summary of Deposits to identify business combinations and branch closings. In our first method, we select mergers with the largest transfer of branches. This is important since the credit shock should be strong enough to affect labor markets – which are typically larger than census tract. For that reason, we restrict to mergers with more than 1000 branches acquired. This leaves us with two specific mergers: the merger of Firstar Corporation with U.S. Bancorp in 2001, and the merger of Bank of America and FleetBoston Financial in 2004. 53

Alternatively, as a form of robustness, we run our analysis using mergers that exactly conform to Nguyen (2018). As she does, we choose mergers where both Buyer and Target held at least $10 billion in premerger assets, and the branch network of each bank overlaps in at least one Census tract. Empirical Specification Nguyen (2018) shows that post-merger branch consolidation reduces local small business lending. In contrast to bank deregulation which occurred at state level, bank mergers led to credit contraction at county levels mostly by limiting access to local branches. Since the effects stemming from bank mergers are more localized, we focus on the effects of credit contractions at the county rather than state level. We can assess whether a reduction in credit increases the gender pay gap in the low-pay-gap industries. In order to do so, again define Ω = {High,Medium,Low} to be the classifications of industries into low, medium, and high preperiod pay gap industries, and Ik is a dummy indicating j whether industry j falls into classification k ∈ Ω. We now have the following specification: (cid:88) (cid:88) (cid:88) Y = α+ β D ×Ik + γ D ×Ik ×F + δ Ik ×F (5) ijct k ct j k ct j i k j i k∈Ω k∈Ω k∈Ω (cid:88) + ζ Ik +πX +τ +µ +(cid:15) k j ijst t,female s,female ijst k∈Ω for D = Post ×Close ct mt cm where i denotes individual, c denotes county, m denotes merger deal and t denotes time. Post mt equals 1 if merger m precedes year t, Close is a dummy equal to 1 if a branch has closed in cm county c after merger m. Effects of Bank Mergers on Gender Pay Gaps We intend to test whether, following weakened credit conditions and absent better job prospects for workers at high-paying high-pay-gap industries, credit-induced relative wage gains for women in the low-pay-gap industries disappears, i.e., relative wages for women would decline. We find that is the case. Table 2 reports effects of bank mergers on wages. While high and median pay gap industries are largely unaffected by bank mergers,thelow-pay-gapindustriesshowareductioninthewagesofwomenofabout3to4%,while wages for men increase by about 2%. All in all, the pay gap increases by about 6%. Importantly, workers in the high-pay-gap industries are unaffected. The results are robust to the inclusion of controls including age, race, and marital status. Jointly, our results so far show that credit expansions alter workers’ calculus of industry choice in a gendered way. However, our bank merger analysis highlights that this effect is not permanent. Credit contractions can make disappear the gains female workers had obtained in the low-paygap industries while not affecting the gains male workers enjoyed in the high-pay-gap industries. Consequently, the emergence of labor dynamics leave women more vulnerable to deterioration of economic conditions. Vulnerability of women’s wages goes hand in hand with changes in the cyclicality of women’s employment. Sincethe1991recession,femaleemploymentcyclicalityhasstartedtoresemblethatof male’s (Albanesi 2019). Moreover, female labor participation has been associated with increases in total factor productivity, while reduced female participation growth (which would follow declines in femalewages)isconnectedwithjoblessrecoveries,affectingoveralleconomicperformance(Albanesi 2019). 54

F. Robustness of Industry Equitability Categorization A potential concern is that the low-pay-gap or high-pay-gap classifications are endogenous outcomes, and thus we cannot include the always-treated states in our analysis. For our main categorization, whereby industries are categorized during the 5 year window spanning 1976 to 1980, there are 17 always-treated states for intrastate deregulation and one always-treated state for interstate deregulation (Maine).21 To mitigate this concern, we show that excluding all seventeen always treated states does not change industry categorization. Appendix Table (F.3) shows that all the high-pay-gap industries remainclassifiedashigh-pay-gapafterexcludingalwaystreatedstates. Onlyoneindustryclassified as low-pay-gap was reclassified after excluding the always treated states: Lumber and building material retailing (CPS ind1990 = 580) moved from the low-pay-gap category to the medium-paygapcategory. Overall, onlytwoindustrieschangedclassification—theotherbeingElectriclightand power (CPS ind1990 = 450) which moved from the medium-pay-gap category to the high-pay-gap category. Tofurthermitigateanyconcerns,wehaveprovidedthreeadditionalsetsofrobustnessanalyses: (1) estimates from both the interstate deregulation and the intrastate deregulation for comparison; (2) estimates using a categorization whereby industries are categorized during the 5 year window spanning 1968 to 1972 (Appendix Table J.10), which reduces always-treated states to 13; and (3) categorization using industry measures of asset tangibility (Table J.11). All estimates are similar in direction, magnitude, and statistical significance. G. Effects on Direct Lending to Worker G.1 Effects of Bank Deregulation on Gender Differences in Housing and Transportation Onepotentialconcernisthatfinancialderegulationoperatesbydirectlyaffectingtheworkerinstead of operating through the assets of the firm. To mitigate these concerns, we estimate Eq. (1) using household outcomes which would directly benefit from increased access to credit: homeownership, holding a mortgage, car ownership, moving into new dwelling (potentially triggered by relocating for a better job), and transportation time (potentially triggered by commutting to a better job). All these dimensions are potentially affected by financial constraints. We report estimates in Tables (G.5) and (G.6). While it is not clear whether relaxing financial constraints for any of these dimensions would lead to the cross-industry we have documented in the paper, it is reassuring to find no economic or statistically meaningful gender differences following deregulation along any of these dimensions for both intrastate and interstate deregulation. G.2 Effects of Bank Deregulation on Gender Differences in Self-Employment Another potential concern is that financial deregulation may affect self-employment opportunities for women. We can test this directly by estimating Eq. (1) using self-employment as an outcome. Self-employment can become easier, if financial constraints are relaxed, or harder, if relaxing the financial constraints of bigger firms makes it harder for individuals to compete. We report estimates in Table (G.7) by type of self-employment for: (1) all industries, (2) lowpay-gapindustriesonly,and(3)high-pay-gapindustriesonly. PanelAshowsestimatesforintrastate 21. Interstate deregulation estimates excluding Maine presented in Table (F.4). 55

deregulation, while Panel B shows effects for interstate deregulation. Intrastate deregulation does not have an effect on gender differences in self-employment for any of the three industry categories andforanytypeofself-employment. Interstatederegulationdoesnothaveeconomicallymeaningful effects on gender differences in unincorporated self-employment. In contrast, for incorporated self-employment, there are small but statistically significant gender differences in incorporated self-employment of between 0.69 and 1.04%. These effects are mostly driven by lower rates of incorporated self-employment among men than increases among women. Despite this, it is not likely that these gender differences in incorporated self-employment for interstate deregulation are driving our core results since the core results hold for both intrastate and interstate deregulation. H. Routine vs. Nonroutine Workers Benefits after deregulation accrue to nonroutine workers at the expense of routine workers. As we discussed in Section IV, industries with low and high pay gaps follow different business models and have different levels of asset tangibility. Consistent with this fact, wage increases in the highpay-gap industries accrue to nonroutine cognitive workers; while wage increases in the low-pay-gap industries accrue to nonroutine manual workers, as shown in Appendix Table (H.8). I. Controlling for Tobin’s Q, Earnings Volatility and Leverage by Gender InTable(I.9)weshowtherobustnessofthemainfindingswhencontrollingforproxiesforindustrial risk taking and Tobins’ Q by gender. J. Alternative Categorizations In this appendix section, we repeat the main estimates of this paper (Table 4) using alternative ways of categorizing workers. In particular, we categorize industries by (i) using 1968–1972 as the categorization period instead of 1976–1980, (ii) by asset tangibility, or (iii) according to worker skills required in each occupation. We show that our main results do not meaningfully change if we follow an alternative categorization procedure. Further analysis on this robustness exercise is contained in Subsection IV.5. J.1 Analysis with Categorization by Industrial Gender Pay Gap in 1968–1972 (Instead of 1976–1980) The main analysis in the text starts in 1976 because states in the CPS data can only be identified separately starting in the 1977 survey. Less precise state identifiers exist, however, for earlier years. Using these imprecise identifiers we can repeat Table (4) using data starting in 1968, which is the earliestyearwhereworkerscanbeclassifiedintofull-timefull-yearstatus. Theresultsarepresented in Table (J.10). The results are generally similar to those presented in the main text. 56

J.2 Analysis with Categorization by Asset Tangibility K. More on Asset Tangibility L. Unstaggered Difference-in-differences Estimates Throughout the paper we showed difference-in-differences (DiD) estimates of the effects of financial deregulationonthegenderpaygapusingastaggeredtreatmentdesign. Thishasbeenthestandard practice in the literature. Nevertheless, recent papers have shown that in estimating differencein-differences with a staggered design and heterogeneous treatment effects some events might be negatively weighted (Sun and Abraham 2020). We can validate the robustness of our main results using a staggered treatment design by showing that the results we obtain when aligning events by event-time instead of calender-time remain similar. We hereinafter refer to this approach as the “unstaggered” DiD design. Our “unstaggered” DiD design approach is similar to Cengiz et al. (2019). For each deregulation event and an X-year bandwidth around the event year, we create event-specific datasets whereby: 1. Only observations at calendar years that fall within the X-year bandwidth around are kept in the sample; and 2. Observations from other states deregulating within the X-year bandwidth around the event year are excluded. These two conditions mean that for each deregulation event all states that did not experience a regulationchangeserveasacontrolgroup(amorestringentcriterionthatprovidesacleanercontrol group) and, also, that there is no other deregulation event contaminating the estimates. For each event h, we define event-time, T, as years since event h, that is, T = calendar-year t−event-h-year. Notice that for each event, controls also have event-year time T defined relative to event-year h. Pooling all datasets, we run a similar equation to Eq. 1: (cid:88) (cid:88) (cid:88) Y = α+ β D ×Ik + γ D ×Ik ×F + δ Ik ×F (6) ijst k st j k st j i k j i k∈Ω k∈Ω k∈Ω (cid:88) + ζ Ik +πX +τ +µ +ρ +(cid:15) k j ijst T,female s,female h ijst k∈Ω wherenowtimefixedeffects, τ , arenowdefinedbasedonevent-time, T, insteadofcalendar- T,female time, T, while ρ indicates event fixed effects. As we mentioned, by aligning events by event-time h and dropping from the control group all states who had deregulated during the X-year bandwidth, this specification gets closer to the canonical DiD model and avoids the negative weighting of some events. We estimate equation 6 for bandwidths X ∈ {1,3,5} for both intrastate and interstate deregulation events. Table L.14 shows our results. The coefficient on Deregulation × Female for the Low-Pay-Gap Industries, the main estimate of interest, ranges from 0.03 to 0.07, which is in the same direction and of a comparable magnitude to the coefficients reported on Table 4. Another important coefficient of interest, Deregulation × for the High-Pay-Gap Industries, is also on the same direction and statistically significant, although under this more stringent specification the magnitudes are smaller. Other estimates are generally of a similar magnitude to the ones reported in Table 4. 57

Appendix Figures and Tables Figure B.1 Balance in Covariates between Nonderegulated and Deregulated (within a year) States Panel A: Intrastate Deregulation Panel B: Interstate Deregulation Notes: This figure shows balance in covariates between states that have been deregulated (just before the passing of deregulation) and states that have not been deregulated. Normalized differences are computed by subtracting the average of each characteristic by deregulationstatusandthencombiningtheaverages. PanelAillustratesthedifferencesinthecaseofintrastatederegulation,andPanel Billustratesthoseforinterstatederegulation. Tangibility,firmriskiness(volatilityoffirmearnings),Tobins’q,andleverageareobtained from Compustat at the industry level and averaged by worker. Thus, they should be interpreted as workers’ exposure to those industry characteristics. Dataonhoursworked,education,age,experience,%black,and%femalearefromtheCPS.Occupationclassificationsby routine/nonroutineandcognitive/manualarebasedonDictionary of Occupational Titles (DOT).Datacovertheyears1976–2014. 58

Figure B.2 Balance in Covariates’ Trends between Nonderegulated and Deregulated (within a year) States Panel A: Intrastate Deregulation Panel B: Interstate Deregulation Notes: Thisfigureshowsbalanceincovariatesbetweenstatesthathavebeenderegulated(justbeforethepassingofderegulation)andstates that have not been deregulated. Normalized differences are computed by subtracting the average of each characteristic by deregulation status and then combining the averages. Tangibility, firm riskiness (volatility of firm earnings), Tobins’ q, and leverage are obtained from Compustat at the industry level and averaged by worker. Thus, they should be interpreted as workers’ exposure to those industry characteristics. Dataonhoursworked,education,age,experience,%black,and%femalearefromtheCPS.Occupationclassificationsby routine/nonroutineandcognitive/manualarebasedonDictionary of Occupational Titles (DOT).Datacovertheyears1976–2014. 59

Figure C.3 Nonroutine/routine and Cognitive/Manual Task in Low and High-Pay-Gap Industries Panel A: Nonroutine Cognitive vs. Nonroutine Manual Task Panel B: Routine Cognitive vs. Routine Manual Task Notes: This figure plots the share of workers performing nonroutine/routine and cognitive/manual tasks computed using the DOT measures developed and defined by Autor, Levy, and Murnane (2003) and the CPS data from 1976-2014. The sample includes full time working-ageadults. ThesampleexcludesindividualsworkingintheFinance,InsuranceandRealEstate(FIRE)industries. Industriesare categorized into the low-pay-gap and high-pay-gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. Formoredetails,seeSectionII.5. 60

Figure C.4 Working Hours by Gender Panel A: Average Working Hours by Male Workers Panel B: Average Working Hours by Female Workers Notes: Thisfigureplotstheaverageweeklyhoursworkedbygenderandindustryduring1980–2010usingtheCPSdata. Thetoppanel plotstheaverageweeklyhoursworkedforfulltimeworking-agemaleemployeesinindustriesexcludingFinance,Insurance,andRealEstate (FIRE)industries. Thebottompanelplotstheaverageweeklyhoursworkedforfemaleemployees. Industriesarecategorizedintolowpay gapandhighpaygapbasedonthedifferenceinthemeanlogwagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustriesthatbelongtothetop25%ofthepaygapdistribution,andthelow-pay-gapindustriesrefer tothoseinthebottom25%ofthepaygapdistribution. 61

Figure C.5 Employment Composition of Low and High-Pay-Gap Industries Panel A: Female Share in Low and High-Pay-Gap Industries Panel B: Share of Hourly-Paid Positions by Gender and Industry Notes: This figure plots the fraction of hourly-paid positions between 1990 and 2014 using the CPS data from 1976-2014. The sample includes full time working-age adults. The sample excludes individuals working in the Finance, Insurance and Real Estate (FIRE) industries. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlogwagebetweenmaleand female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, and the low-pay-gap industries refer to those in the bottom 25% of the pay gap distribution. For more details, see SectionII.5. 62

Table D.1: Oaxaca Blinder Decomposition Pre- and Post-Deregulation YearPre-Deregulation FiveYearsPost-Deregulation Difference LogPoints Percentage LogPoints Percentage LogPoints Percentage (1) (2) (3) (4) (5) (6) TotalPayGap 0.423 100.0% 0.312 100% -0.111 100% HighPayGapIndustry 0.034 8.0 0.022 7.0 -0.012 0.107 LowPayGapIndustry -0.013 -3.1 -0.039 -12.5 -0.026 0.237 Notes: ThistablereportstheOaxaca-Blinderestimatesofbankderegulationonthepaygapforfull-timefull-yearworkers,excludingthe Finance, Insurance, and Real Estate (FIRE) industries. Columns (1)–(2) show estimates calculated for the year immediately preceding deregulationinthestate. Columns(3)–(4)showestimatescalculatedfiveyearsfollowingderegulation. Columns(5)–(6)showthedifference. Wages in the wage regressions are the residual of a regression of log wages on Mincerian traits (education, experience, and experience squared)byyear,andyearandstatefixedeffects. Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferencein themeanlogwagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustries that belong to the top 25% of the pay gap distribution, and the low-pay-gap industries refer to those in the bottom 25% of the pay gap distribution. 63

Table E.2: Effects of Bank Mergers on Gender Pay Gap (1) (2) (3) (4) Merger×Female 0.02 0.02 0.02 0.02 (0.01) (0.01) (0.01) (0.01) Merger×Female–LowPGIndustry -0.06∗∗∗ -0.06∗∗∗ -0.06∗∗∗ -0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) Merger×Female–HighPGIndustry 0.01 0.01 0.01 0.01 (0.01) (0.01) (0.01) (0.01) Merger -0.01 -0.01 -0.01 -0.01 (0.01) (0.01) (0.01) (0.01) Merger–LowPGIndustry 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗ (0.01) (0.01) (0.01) (0.01) Merger–HighPGIndustry 0.01 0.01 0.01 0.01 (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.12∗∗∗ 0.12∗∗∗ 0.12∗∗∗ 0.12∗∗∗ (0.00) (0.00) (0.00) (0.00) Female–HighPGIndustry -0.03∗∗∗ -0.03∗∗∗ -0.02∗∗∗ -0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.02∗∗∗ -0.02∗∗∗ -0.02∗∗∗ -0.02∗∗∗ (0.00) (0.00) (0.00) (0.00) HighPGIndustry 0.13∗∗∗ 0.12∗∗∗ 0.12∗∗∗ 0.12∗∗∗ (0.00) (0.00) (0.00) (0.00) N 477,550 477,550 477,550 477,550 County×Gender Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Age×Gender No Yes Yes Yes MaritalStatus×Gender No No Yes No Age×Gender No No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsthedifference-in-differencesestimatesoftheeffectsofbankmergeronthegenderpaygapwhenlog(hourlywage) isregressedonasetofindicatorsandcontrols,asspecifiedinEquation(2). Industriesarecategorizedintolowpaygapandhighpaygap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industriesrefertoindustriesthatbelongtothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthe bottom25%ofthepaygapdistribution. AllspecificationscontrolforMinceriantraits×gender,andcounty×genderandyear×genderfixed effects. Columns (2)–(4) and (6)–(8) additionally control for age×gender fixed effects. For details, see Section E. Errors are clustered at thecountylevelandreportedinparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 64

Table F.3: Comparison of Industry Categorization using Alternative Sample — Excluding States Always-Treated for Intrastate Bank Deregulation #IndustriesinSubsample #IndustriesUnchangedAfterRecategorization MatchRate(%) OriginalCategorization ExcludingAlwaysTreated (1) (2) (3) Panel A: All Industries 189 187 99% Panel B: Low Pay Gap Industries 46 45 98% Panel C: High Pay Gap Industries 51 51 100% Notes: Thetablereportsthenumberoflowandhigh-pay-gapindustrieswithinasubsampleexcludingthe17statesthatderegulatedprior to 1980. Column (1) shows the number of total, low-, and high-pay-gap industries categorized using the full sample. Column (2) shows thenumberofindustrieswhosecategoriesremainunchangedaftertheyarerecategorizedintolow-,medium-,andhigh-pay-gapindustries using the subsample. Column (3) reports the match rate between the main and sub-sample. Two industries changed categories after re-categorization: Electric light and power (CPS ind1990 = 450) moved from the medium-pay-gap to the high-pay-gap category, while Lumber and building material retailing (CPSind1990=580)movedfromthelow-pay-gaptothemedium-pay-gapcategory. 65

Table F.4: Effects of Interstate Bank Deregulation on Gender Pay Gap — Excluding States Always-Treated for Intrastate Bank Deregulation (1) (2) (3) (4) Deregulation×Female -0.02∗∗ -0.02∗∗ -0.02∗∗ -0.02∗∗ (0.01) (0.01) (0.01) (0.01) Deregulation×Female–LowPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.04∗∗∗ 0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighPGIndustry 0.01 0.01 0.01 0.01 (0.01) (0.01) (0.01) (0.01) Deregulation -0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) Deregulation–LowPGIndustry 0.01∗ 0.01∗ 0.02∗∗ 0.01∗ (0.01) (0.01) (0.01) (0.01) Deregulation–HighPGIndustry 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ (0.01) (.01) (.01) (.01) LowPGIndustry -0.19∗∗∗ -.19∗∗∗ -.19∗∗∗ -.19∗∗∗ (0.01) (.01) (.01) (.01) HighPGIndustry 0.02∗∗ 0.02∗∗ 0.02∗∗∗ 0.02∗∗ (0.01) (0.01) (0.01) (0.01) N 804,878 804,878 804,878 804,878 State×Gender Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Age×Gender No Yes Yes Yes MaritalStatus×Gender No No Yes No Race×Gender No No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsthedifference-in-differencesestimatesoftheeffectsofbankderegulationonthegenderpaygapwhenlog(hourly wage)isregressedonasetofindicatorsandcontrols,asspecifiedinEquation(1),excludingstatesthatderegulatedpriorto1980. Columns (1)–(4) report the effects of intrastate deregulation, excluding the 17 states that deregulated prior to 1980. Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer to industries that belong to the top 25% of the pay gap distribution, and the lowpay-gap industries refer to those in the bottom 25% of the pay gap distribution. All specifications control for Mincerian traits×gender, andstate×genderandyear×genderfixedeffects. Columns(3)–(4)additionallycontrolforage×genderfixedeffects. Formoredetails,see Section II.3. Errors are clustered at the state level and reported in parentheses. *,**, and *** indicate significance at the 10%, 5%, and 1%levels,respectively. 66

Table G.5: Effects of Intrastate Bank Deregulation on Gender Differences in Housing and Transportation OwnsHouse MovedHouse Mortgage OwnsCar TransportationTime (1) (2) (3) (4) (5) Panel A: All Industries Deregulation×Female -0.0024 -0.0006 -0.0014 -0.0015 0.0069 (0.0068) (0.0051) (0.0024) (0.0032) (0.0063) Deregulation 0.0171∗ -0.0035 -0.0102 0.0153∗∗ -0.0032 (0.0092) (0.0065) (0.0082) (0.0067) (0.0147) N 815,650 688,547 5,345,055 8,806,388 6,144,008 Panel B: Low Pay Gap Industries Deregulation×Female -0.0088 0.0015 -0.0036 -0.0064 0.0072 (0.0097) (0.0093) (0.0029) (0.0042) (0.0078) Deregulation 0.0181∗∗ -0.0052 -0.0085 0.0150∗∗ -0.0063 (0.0090) (0.0085) (0.0072) (0.0064) (0.0119) N 207,486 179,480 1,139,255 1,972,398 1,412,705 Panel C: High Pay Gap Industries Deregulation×Female 0.0051 0.0041 -0.0000 -0.0003 -0.0015 (0.0100) (0.0107) (0.0046) (0.0027) (0.0076) Deregulation 0.0063 -0.0052 -0.0060 0.0152∗∗ 0.0099 (0.0107) (0.0084) (0.0092) (0.0063) (0.0148) N 205,400 172,006 1,279,888 2,058,252 1,421,266 County×Gender Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Controls Yes Yes Yes Yes Yes Data CPS CPS Census Census Census ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of the effects of intrastate bank deregulation on differences in housing and transportation by gender using the CPS data from 1976–2014 and the Census data from 1980–2000. Both samples are restricted to working-age full-time full-year workers in the private sectors, excluding the FIRE industries. The dependent variables are ownership of dwellingforcolumn(1),movingtoadifferenthouseforcolumn(2),holdingamortgageforcolumn(3),carownershipforcolumn(4),and transportationtimeforcolumn(5). Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlog wagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustriesthatbelong tothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. AllspecificationscontrolforMinceriantraits×gender,andstate×genderandyear×genderfixedeffects. Formoredetails,seeSectionII.3. Errors are clustered at the state level and reported in parentheses. *,**, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 67

Table G.6: Effects of Interstate Bank Deregulation on Gender Differences in Housing and Transportation OwnsHouse MovedHouse Mortgage OwnsCar TransportationTime (1) (2) (3) (4) (5) Panel A: All Industries Deregulation×Female -0.0014 -0.0000 -0.0118 -0.0053 0.0053 (0.0025) (0.0030) (0.0117) (0.0072) (0.0075) Deregulation -0.0014 0.0014 -0.0135 0.0114∗ -0.0022 (0.0067) (0.0055) (0.0139) (0.0062) (0.0048) N 5,345,055 8,806,388 6,144,008 815,650 688,547 Panel B: Low Pay Gap Industries Deregulation×Female -0.0074 -0.0034 -0.0064 -0.0187 -0.0053 (0.0097) (0.0049) (0.0189) (0.0180) (0.0114) Deregulation -0.0079 -0.0011 -0.0070 0.0238∗∗ -0.0071 (0.0078) (0.0072) (0.0160) (0.0093) (0.0118) N 1,139,255 1,972,398 1,412,705 207,486 179,480 Panel C: High Pay Gap Industries Deregulation×Female -0.0015 -0.0029 -0.0155 -0.0169 0.0093 (0.0041) (0.0034) (0.0112) (0.0128) (0.0146) Deregulation 0.0012 0.0013 -0.0206∗∗∗ 0.0119 -0.0084 (0.0083) (0.0035) (0.0040) (0.0073) (0.0063) N 1,279,888 2,058,252 1,421,266 205,400 172,006 County×Gender Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Controls Yes Yes Yes Yes Yes Data CPS CPS Census Census Census ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of the effects of interstate bank deregulation on differences in housing and transportation by gender using the CPS data from 1976–2014 and the Census data from 1980–2000. Both samples are restricted to working-age full-time full-year workers in the private sectors, excluding the FIRE industries. The dependent variables are ownership of dwellingforcolumn(1),movingtoadifferenthouseforcolumn(2),holdingamortgageforcolumn(3),carownershipforcolumn(4),and transportationtimeforcolumn(5). Industriesarecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlog wagebetweenmaleandfemaleemployeesineachindustryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustriesthatbelong tothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. AllspecificationscontrolforMinceriantraits×gender,andstate×genderandyear×genderfixedeffects. Formoredetails,seeSectionII.3. Errors are clustered at the state level and reported in parentheses. *,**, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 68

tnemyolpmE-fleS ni secnereffiD redneG no noitalugereD knaB fo stceffE :7.G elbaT deyolpmE-fleS|detaroprocnI detaroprocninUdeyolpmE-fleS detaroprocnIdeyolpmE-fleS paGyaPhgiH paGyaPwoL llA paGyaPhgiH paGyaPwoL llA paGyaPhgiH paGyaPwoL llA )9( )8( )7( )6( )5( )4( )3( )2( )1( noitalugereD etatsartnI :A lenaP 6900.0- 1900.0 9000.0 4000.0 5000.0 1000.0 5600.0 9200.0 4300.0 elameF×noitalugereD )1930.0( )3440.0( )8820.0( )2100.0( )1100.0( )6000.0( )9300.0( )5300.0( )9200.0( 2500.0- 3300.0- 1400.0- 4100.0- 3000.0- 2000.0- 7200.0- 5000.0- 2100.0noitalugereD )6810.0( )3220.0( )3810.0( )8100.0( )9000.0( )9000.0( )5500.0( )2400.0( )3400.0( 763,71 912,11 346,15 816,072 682,092 630,412,1 816,072 682,092 630,412,1 N noitalugereD etatsretnI :B lenaP 5130.0 6630.0- 9100.0 4000.0- ∗∗7200.0 4000.0 ∗∗1800.0 ∗∗∗4010.0 ∗∗∗9600.0 elameF×noitalugereD )4750.0( )1050.0( )0120.0( )0100.0( )2100.0( )4000.0( )1300.0( )5200.0( )6100.0( 6410.0- 2030.0- ∗6220.0- 2100.0 1100.0- 5000.0 6500.0- ∗∗∗1010.0- ∗∗∗8600.0noitalugereD )1610.0( )7040.0( )5210.0( )1100.0( )2100.0( )5000.0( )4400.0( )8200.0( )4200.0( 763,71 912,11 346,15 816,072 682,092 630,412,1 816,072 682,092 630,412,1 N seY seY seY seY seY seY seY seY seY redneG×ytnuoC seY seY seY seY seY seY seY seY seY redneG×raeY seY seY seY seY seY seY seY seY seY slortnoC 10.0<p ∗∗∗ ,50.0<p ∗∗ ,01.0<p ∗ gnisu)1(noitauqEgniwolloftnemyolpme-flesnisecnereffidrednegehtnonoitalugeredknabfostceffeehtfosetamitsesecnereffid-ni-ecnereffidstroperelbatsihT :setoN gnidulcxe rotces etavirp eht ni slaudividni deyolpme-fles dna srekrow raey-lluf emit-lluf ega-gnikrow ot tcirtser eW .4102–6791 morf )SPC( yevruS noitalupoP tnerruC elbairavrotacidninasi)6(–)4(snmulocrof ;sutatsdetaroprocnideyolpme-flesrofelbairavrotacidninasi)3(–)1(snmulocrofelbairavtnednepedehT .seirtsudniERIF seirtsudnI .tnemyolpme-flesnolanoitidnocsutatsdetaroprocnideyolpme-flesrofelbairavrotacidninasi)9(–)7(snmulocrofdna;sutatsdetaroprocninudeyolpme-flesrof .0891–6791gnirudyrtsudnihcaeniseeyolpmeelamefdnaelamneewtebegawgolnaemehtniecnereffidehtnodesabpagyaphgihdnapagyapwolotnidezirogetacera %52mottobehtniesohtotreferseirtsudnipag-yap-wolehtdna,noitubirtsidpagyapehtfo%52potehtotgnolebtahtseirtsudniotreferseirtsudnipag-yap-hgihehT noitceS ees ,sliated lanoitidda roF .stceffe dexfi redneg×raey dna redneg×etats dna ,redneg×stiart nairecniM rof lortnoc snoitacfiiceps llA .noitubirtsid pag yap eht fo .ylevitcepser,slevel%1dna,%5,%01ehttaecnacfiingisetacidni***dna,**,* .sesehtnerapnidetroperdnaleveletatsehttaderetsulcerasrorrE .3.II 69

Table H.8: Effects of Bank Deregulation on Gender Pay Gap by Worker Skill IntrastateDeregulation InterstateDeregulation (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -.01 -.00 -.00 -.00 -.01 -.01 -.01 -.01 (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Deregulation×Female–NRManualOccupation .05∗∗∗ .05∗∗∗ .05∗∗∗ .05∗∗∗ .08∗∗∗ .08∗∗∗ .08∗∗∗ .09∗∗∗ (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Deregulation×Female–NRCognitiveOccupation .01 .01 .01 .01 .01 .01 .01 .01 (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Deregulation -.05∗∗∗ -.06∗∗∗ -.06∗∗∗ -.06∗∗∗-.07∗∗∗ -.07∗∗∗ -.07∗∗∗ -.07∗∗∗ (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Deregulation–NRManualOccupation -.03∗∗∗ -.02∗∗∗ -.03∗∗∗ -.03∗∗∗-.04∗∗∗ -.04∗∗∗ -.04∗∗∗ -.04∗∗∗ (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Deregulation–NRCognitiveOccupation .11∗∗∗ .15∗∗∗ .14∗∗∗ .15∗∗∗ .16∗∗∗ .16∗∗∗ .15∗∗∗ .16∗∗∗ (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Female–NRManualOccupation -.02 -.07∗∗∗ -.06∗∗∗ -.06∗∗∗-.09∗∗∗ -.09∗∗∗ -.09∗∗∗ -.09∗∗∗ (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Female–NRCognitiveOccupation .00 .00 .01 .01 -.01 -.01 -.00 -.01 (.01) (.01) (.01) (.01) (.01) (.01) (.01) (.01) NRManualOccupation .00 .03∗∗∗ .03∗∗∗ .03∗∗∗ .04∗∗∗ .04∗∗∗ .04∗∗∗ .04∗∗∗ (.) (.01) (.01) (.01) (.01) (.01) (.01) (.01) NRCognitiveOccupation .00 .14∗∗∗ .14∗∗∗ .14∗∗∗ .14∗∗∗ .14∗∗∗ .14∗∗∗ .14∗∗∗ (.) (.01) (.01) (.01) (.01) (.01) (.01) (.01) Black -.12∗∗∗ -.12∗∗∗ (.01) (.01) Married .15∗∗∗ .15∗∗∗ (.00) (.00) N 812716 812716 812716 812716812716 812716 812716 812716 County×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender No Yes Yes Yes No Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsdifferences-in-differencesestimatesoftheimpactofbankderegulationongenderpaygapregressinglog(hourly wage) on a set of indicators and controls specified in Equation (1) but categorizing workers by nonroutine cognitive, nonroutine manual, andallroutineoccupations. Columns(1)–(4)reportstheimpactofintrastatederegulationasatreatment,andcolumns(5)–(8)reportsthe impactofinterstatederegulationasatreatment. AllspecificationscontrolforMinceriantraits×gender,andstate×genderandyear×gender fixed effects. Columns (2), (4), (6) and (8) additionally control for occupation×gender fixed effects, while columns (3)–(4) and (7)–(8) control for industry×gender fixed effects. For additional details, see Section II.3. Errors clustered at the state level and reported in parentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 70

Table I.9: Effects of Bank Deregulation on Gender Pay Gap — Additional Industry-Level Controls IntrastateDeregulation InterstateDeregulation (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -0.01 -0.01 -0.02∗ -0.01 -0.02∗ -0.01 -0.02∗ -0.02∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–LowPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.06∗∗∗ 0.05∗∗∗ 0.04∗∗∗ 0.04∗∗∗ 0.05∗∗∗ 0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighPGIndustry -0.01 -0.01 -0.00 -0.01 -0.00 -0.00 0.00 0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗-0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ -0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–LowPGIndustry 0.02 0.01 0.01 0.01 0.04∗∗∗ 0.03∗∗∗ 0.03∗∗∗ 0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–HighPGIndustry 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.10∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.09∗∗∗ 0.03∗∗ 0.10∗∗∗ 0.03∗∗ 0.10∗∗∗ 0.04∗∗∗ 0.11∗∗∗ 0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.04∗∗∗ -0.05∗∗∗ -0.04∗∗∗ -0.06∗∗∗-0.05∗∗∗ -0.06∗∗∗ -0.04∗∗∗ -0.07∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.19∗∗∗ -0.17∗∗∗ -0.18∗∗∗ -0.18∗∗∗-0.20∗∗∗ -0.18∗∗∗ -0.19∗∗∗ -0.20∗∗∗ (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) HighPGIndustry 0.01 0.02∗ 0.01 0.00 0.00 0.01 0.00 -0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female×Tobins’Q -0.05∗∗ 0.08∗∗∗ -0.04∗∗ 0.08∗∗∗ (0.02) (0.02) (0.02) (0.02) Female×Leverage 0.24∗∗∗ 0.23∗∗∗ 0.24∗∗∗ 0.23∗∗∗ (0.03) (0.03) (0.03) (0.03) Female×EarningsVolatility 0.00∗∗∗ 0.00∗∗∗ 0.00∗∗∗ 0.00∗∗∗ (0.00) (0.00) (0.00) (0.00) Tobins’Q 0.10∗∗∗ 0.11∗∗∗ 0.10∗∗∗ 0.11∗∗∗ (0.02) (0.02) (0.02) (0.02) Leverage 0.17∗∗∗ 0.23∗∗∗ 0.17∗∗∗ 0.23∗∗∗ (0.02) (0.03) (0.02) (0.03) EarningsVolatility 0.00∗∗∗ 0.00∗∗∗ 0.00∗∗∗ 0.00∗∗∗ (0.00) (0.00) (0.00) (0.00) N 711,241 711,241 711,241 711,241 711,241 711,241 711,241 711,241 County×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender Yes Yes Yes No No Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsdifference-in-differencesestimatesoftheeffectsofbankderegulationongenderpaygapwhenlog(hourlywage)is regressedonthesetofindicatorsandcontrolsspecifiedinEquation(1)andindustry-levelcontrolsareincluded. Columns(1)–(4)reportthe effectsofintrastatederegulationasatreatment,andcolumns(5)–(8)reporttheeffectsofinterstatederegulationasatreatment. Industries arecategorizedintolowpaygapandhighpaygapbasedonthedifferenceinthemeanlogwagebetweenmaleandfemaleemployeesineach industryduring1976–1980. Thehigh-pay-gapindustriesrefertoindustriesthatbelongtothetop25%ofthepaygapdistribution,andthe low-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. AllspecificationscontrolforMinceriantraits×gender, andstate×genderandyear×genderfixedeffects. Foradditionaldetails,seeSectionII.3. Errorsareclusteredatthestatelevelandreported inparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 71

Table J.10: Effects of Bank Deregulation on Gender Pay Gap — Industries Categorized based on Pay Gaps from 1968 to 1972 IntrastateDeregulation InterstateDeregulation (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -0.02 -0.02 -0.02 -0.02 -0.02 -0.02 -0.01 -0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–LowPGIndustry 0.06∗∗∗ 0.06∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighPGIndustry 0.01 0.01 0.01 0.01 0.02∗∗ 0.02∗∗ 0.01∗ 0.02∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation -0.02 -0.02 -0.02 -0.02 -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–LowPGIndustry -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗ -0.04∗∗∗-0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–HighPGIndustry 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.06∗∗∗ 0.06∗∗∗ 0.06∗∗∗ 0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.06∗∗∗ 0.06∗∗∗ 0.06∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.02∗∗ -0.02∗∗ -0.02∗∗ -0.02∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ -0.03∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.11∗∗∗ -0.11∗∗∗ -0.11∗∗∗ -0.11∗∗∗-0.12∗∗∗ -0.12∗∗∗ -0.12∗∗∗ -0.12∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) HighPGIndustry 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Black -0.12∗∗∗ -0.12∗∗∗ (0.01) (0.01) Married 0.16∗∗∗ 0.16∗∗∗ (0.00) (0.00) N 774,186 774,186 774,186 774,186 774,186 774,186 774,186 774,186 County×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender No Yes Yes Yes No Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of the effects of bank deregulation on gender pay gap when log(hourly wage) is regressed on the set of indicators and controls specified in Equation (1) and industries are categorized based on the gender pay gapduringtheyears1968–1972insteadof1976–1980. Columns(1)–(4)reportstheimpactofintrastatederegulationasatreatment,and columns(5)–(8)reportstheimpactofinterstatederegulationasatreatment. Industriesarecategorizedintolowpaygapandhighpaygap based on the difference in the mean log wage between male and female employees in each industry during 1968-1972. The high-pay-gap industriesrefertoindustriesthatbelongtothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthe bottom25%ofthepaygapdistribution. AllspecificationscontrolforMinceriantraits×gender,andstate×genderandyear×genderfixed effects. Columns(2)–(4)and(6)–(8)additionallycontrolforage×genderfixedeffects. Fordetails,seeSectionII.3. Errorsareclusteredat thestatelevelandreportedinparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 72

Table J.11: Effects of Bank Deregulation on Gender Pay Gap, — Industries Categorized by Asset Tangibility IntrastateDeregulation InterstateDeregulation (1) (2) (3) (4) (5) (6) (7) (8) Deregulation×Female -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation×Female–HighTangibility 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.03∗ 0.03∗ 0.03∗ 0.03∗ (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Deregulation×Female–LowTangibility 0.00 0.00 0.00 0.00 0.01 0.01 0.01 0.02 (0.02) (0.02) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) Deregulation -0.01 -0.01 -0.02 -0.02 -0.02∗ -0.02∗ -0.02∗∗ -0.02∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Deregulation–HighTangibilityIndustry -0.05∗∗∗ -0.05∗∗∗ -0.05∗∗ -0.06∗∗∗-0.04∗∗∗ -0.04∗∗∗ -0.03∗∗ -0.04∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) Deregulation–LowTangibilityIndustry 0.07∗∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.06∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighTangibilityIndustry -0.09∗∗∗ -0.09∗∗∗ -0.09∗∗∗ -0.08∗∗∗ -0.05∗∗ -0.05∗∗ -0.05∗∗ -0.05∗∗ (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Female–LowTangibilityIndustry 0.02 0.02 0.01 0.01 0.00 0.00 0.00 -0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) HighTangibilityIndustry -0.14∗∗∗ -0.14∗∗∗ -0.14∗∗∗ -0.14∗∗∗-0.16∗∗∗ -0.16∗∗∗ -0.15∗∗∗ -0.16∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) LowTangibilityIndustry -0.10∗∗∗ -0.10∗∗∗ -0.09∗∗∗ -0.09∗∗∗-0.08∗∗∗ -0.08∗∗∗ -0.08∗∗∗ -0.08∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Black -0.14∗∗∗ -0.14∗∗∗ (0.01) (0.01) Married 0.16∗∗∗ 0.16∗∗∗ (0.00) (0.00) N 867,993 867,993 867,993 867,993 867,993 867,993 867,993 867,993 County×Gender Yes Yes Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes Yes Yes Age×Gender No Yes Yes Yes No Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsthedifference-in-differencesestimatesoftheeffectsofbankderegulationonthegenderpaygapwhenlog(hourly wage) is regressed on a set of indicators and controls specified, as in Equation (1), and industries are categorized by their level of asset tangibility. Columns(1)–(4)reporttheeffectsofintrastatederegulationasatreatment,andcolumns(5)–(8)reporttheeffectsofinterstate deregulationasatreatment. Industriesarecategorizedintolowassettangibilityandhighassettangibilitybasedonthedifferenceinthe meanassettangibilityshareineachindustryduring1976–1980. Thehigh-asset-tangibilityindustriesrefertoindustriesthatbelongtothe top25%oftheassettangibilitydistribution,andthelow-asset-tangibilityindustriesrefertothoseinthebottom25%oftheassettangibility distribution. Hightangibilityisadummyvariablethattakesthevalueoneforthehigh-asset-tangibilityindustriesand0otherwise. Low tangibilityisadummyvariablethattakesthevalueoneforthelow-asset-tangibilityindustriesand0otherwise. Allspecificationscontrolfor Minceriantraits×gender,andstate×genderandyear×genderfixedeffects. Columns(2)–(4)and(6)–(8)additionallycontrolforage×gender fixedeffects. Formoredetails,seeSectionII.3. Errorsareclusteredatthestatelevelandreportedinparentheses. *,**,and***indicate significanceatthe10%,5%,and1%levels,respectively. 73

Table K.12: Effects of Deregulation on Firm Borrowing — Industries Categorized by Asset Tangibility DebtRatio DebtGrowth LongTermDebtGrowth (1) (2) (3) (4) (5) (6) (7) (8) (9) Intrastate–LowTangibility -0.10 -0.08 -0.06 -0.00 -0.00 -0.01 0.01 0.01 0.01 (0.09) (0.09) (0.09) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Intrastate–HighTangibility 0.14∗∗∗ 0.18∗∗∗ 0.18∗∗∗ 0.05∗∗ 0.03 0.03 0.05∗∗∗ 0.03 0.04∗ (0.05) (0.05) (0.05) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Intrastate 0.01 0.00 0.00 0.03 0.00 0.00 0.02 0.00 0.00 (0.03) (0.) (0.) (0.02) (0.) (0.) (0.02) (0.) (0.) TotalAssets 0.11∗∗∗ 0.11∗∗∗ 0.11∗∗∗ 0.22∗∗∗ 0.22∗∗∗ 0.22∗∗∗ 0.18∗∗∗ 0.18∗∗∗ 0.19∗∗∗ (0.03) (0.03) (0.03) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Tobin’sQ -0.00 -0.00 0.01 (0.04) (0.02) (0.01) BookLeverage N 61,612 61,553 60,551 61,574 61,515 60,515 61,621 61,562 60,547 Interstate–LowTangibility -0.07 -0.04 -0.04 0.00 0.01 0.01 0.01 0.02 0.02 (0.06) (0.06) (0.06) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Interstate–HighTangibility 0.12∗∗∗ 0.15∗∗∗ 0.14∗∗∗ 0.03 0.02 0.02 0.04 0.02 0.02 (0.04) (0.04) (0.04) (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) Interstate 0.02 0.00 0.00 0.04∗ 0.00 0.00 0.04 0.00 0.00 (0.03) (0.) (0.) (0.02) (0.) (0.) (0.03) (0.) (0.) TotalAssets 0.11∗∗∗ 0.11∗∗∗ 0.11∗∗∗ 0.22∗∗∗ 0.22∗∗∗ 0.22∗∗∗ 0.18∗∗∗ 0.18∗∗∗ 0.19∗∗∗ (0.03) (0.03) (0.03) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Tobin’sQ -0.00 -0.00 0.01 (0.04) (0.02) (0.01) BookLeverage N 61,612 61,553 60,551 61,574 61,515 60,515 61,621 61,562 60,547 FirmFX Yes Yes Yes Yes Yes Yes Yes Yes Yes YearFX Yes Yes Yes Yes Yes Yes Yes Yes Yes StateFX Yes Yes Yes Yes Yes Yes Yes Yes Yes State×YearFX No Yes Yes No Yes Yes No Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: Thistablereportsestimatesoftheeffectsofbankderegulationonfirmdebt. Thedependentvariableisdebtratioincolumns(1)– (3),debtgrowthincolumns(4)–(6),long-termdebtgrowthcolumns(7)–(9). Industriesarecategorizedintolowassettangibilityandhigh assettangibilitybasedonthedifferenceinthemeanassettangibilityshareineachindustryduring1976–1980. Thehigh-asset-tangibility industriesrefertoindustriesthatbelongtothetop25%oftheassettangibilitydistribution,andthelow-asset-tangibilityindustriesrefer to those in the bottom 25% of the asset tangibility distribution. High tangibility is a dummy variable that takes the value one for the high-asset-tangibilityindustriesand0otherwise. Lowtangibilityisadummyvariablethattakesthevalueoneforthelow-asset-tangibility industries and 0 otherwise. All specifications control for firms, state, and year fixed effects. Columns (2), (5), (8), and (11) additionally controlforstate×yearfixedeffects. Fordetails,seeSectionII.3. Errorsareclusteredatthestatelevelandreportedinparentheses. *,**, and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 74

Table K.13: Effects of Deregulation on Firm Percentage of Tangible Assets %Tangible R&DSpending (1) (2) (3) (4) (5) (6) Intrastate–HighPGIndustry -0.00 -0.01 -0.01 0.31∗∗∗ 0.32∗∗∗ 0.32∗∗∗ (0.00) (0.01) (0.00) (0.06) (0.05) (0.03) Intrastate–LowPGIndustry 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗∗ -0.09 0.01 0.09 (0.01) (0.01) (0.01) (0.08) (0.09) (0.07) Intrastate -0.00 -0.29∗∗∗ (0.01) (0.06) N 68,407 68,355 60,593 41,535 41,387 36,541 Interstate–HighPGIndustry -0.01∗∗ -0.01∗∗ -0.01∗∗∗ 0.35∗∗∗ 0.33∗∗∗ 0.29∗∗∗ (0.00) (0.00) (0.00) (0.05) (0.04) (0.03) Interstate–LowPGIndustry 0.02∗∗∗ 0.02∗∗∗ 0.01∗∗∗ -0.12∗ -0.07 0.01 (0.00) (0.01) (0.01) (0.07) (0.08) (0.06) Interstate -0.00 -0.22∗∗∗ (0.00) (0.04) N 68,407 68,355 60,593 41,535 41,387 36,541 FirmFX Yes Yes Yes Yes Yes Yes YearFX Yes Yes Yes Yes Yes Yes StateFX Yes Yes Yes Yes Yes Yes State×YearFX No Yes Yes No Yes Yes FirmControls No No Yes No No Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports estimates of the effects of bank deregulation on tangible asset share (columns 1–3) and R&D (columns 4–6). Industries are categorized into low pay gap and high pay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. The high-pay-gap industries refer toindustriesthat belong tothe top 25% of the pay gap distribution, andthelow-pay-gapindustriesrefertothoseinthebottom25%ofthepaygapdistribution. HighPGisadummyvariable thattakesthevalueoneforhigh-pay-gapindustriesand0otherwise. LowPGisadummyvariablethattakesthevalueoneforlow-pay-gap industries and 0 otherwise. All specifications control for firms, state, and year fixed effects. Columns (2), (5), (8), and (11) additionally control for state×year fixed effects. Deregulation equals one if intrastate branching is deregulated and zero otherwise. For details, see Section II.3. Errors are clustered at the state level and reported in parentheses. *,**, and *** indicate significance at the 10%, 5%, and 1%levels,respectively. 75

Table L.14: “Unstaggered” DiD Estimates of Deregulation on Gender Pay Gap Intrastate Interstate (1) (2) (3) (4) (5) (6) 1-yearBW 3-yearBW 5-yearBW 1-yearBW 3-yearBW 5-yearBW Deregulation×Female -0.01 -0.01 -0.01 -0.03∗∗ -0.03∗∗ -0.03 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) Deregulation×Female–LowPGIndustry 0.04∗∗∗ 0.03∗∗ 0.03∗∗ 0.05∗∗∗ 0.07∗∗∗ 0.07∗∗∗ (0.01) (0.02) (0.02) (0.02) (0.02) (0.02) Deregulation×Female–HighPGIndustry -0.01 -0.01 -0.01 0.01 0.03∗∗ 0.04∗∗ (0.02) (0.02) (0.02) (0.01) (0.01) (0.02) Deregulation -0.02∗∗ -0.02 -0.02 -0.03∗ -0.02 0.01 (0.01) (0.01) (0.02) (0.01) (0.01) (0.02) Deregulation–LowPGIndustry -0.03∗ -0.02 -0.02 -0.03 -0.02∗ -0.02 (0.01) (0.01) (0.01) (0.02) (0.01) (0.02) Deregulation–HighPGIndustry 0.03∗∗∗ 0.04∗∗∗ 0.03∗∗∗ 0.03∗∗∗ 0.04∗∗∗ 0.05∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–LowPGIndustry 0.11∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.11∗∗∗ 0.10∗∗∗ 0.10∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Female–HighPGIndustry -0.03∗∗∗ -0.03∗∗ -0.03∗∗ -0.05∗∗∗ -0.05∗∗∗ -0.06∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) LowPGIndustry -0.18∗∗∗ -0.19∗∗∗ -0.19∗∗∗ -0.18∗∗∗ -0.18∗∗∗ -0.18∗∗∗ (0.02) (0.01) (0.01) (0.02) (0.02) (0.02) HighPGIndustry 0.04∗∗∗ 0.04∗∗∗ 0.04∗∗∗ 0.05∗∗∗ 0.04∗∗∗ 0.04∗∗∗ (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) N 1,890,824 3,544,025 4,645,609 1,798,185 2,027,712 1,422,063 County×Gender Yes Yes Yes Yes Yes Yes Year×Gender Yes Yes Yes Yes Yes Yes ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01 Notes: This table reports the difference-in-differences estimates of the effects of bank deregulation on gender pay gap when we regress log(hourly wage) on the set of indicators and controls specified in Equation (1) and align events by event-time instead of calender-time (“unstaggered”DiDdesign)similartoCengizetal. (2019). Columns(1)–(3)reporttheeffectsofintrastatederegulationasatreatment, andcolumns(5)–(8)reporttheeffectsofinterstatederegulationasatreatment. Industriesarecategorizedintolowpaygapandhighpay gap based on the difference in the mean log wage between male and female employees in each industry during 1976–1980. High-pay-gap industriesrefertoindustriesthatbelongtothetop25%ofthepaygapdistribution, andthelow-pay-gapindustriesrefertothoseinthe bottom25%ofthepaygapdistribution. HighPGisadummyvariablethattakesthevalueoneforhigh-pay-gapindustriesand0otherwise. LowPGisadummyvariablethattakesthevalueoneforlow-pay-gapindustriesand0otherwise. AllspecificationscontrolforMincerian traits×gender,andstate×genderandyear×genderfixedeffects. Columns(1)and(4),(2)and(5),and(3)and(6)useabandwith(years around deregulation event) of 1, 3, and 5 years, respectively. For additional details, see Section L. Errors are clustered at the state level andreportedinparentheses. *,**,and***indicatesignificanceatthe10%,5%,and1%levels,respectively. 76