Large-scale Evidence Research Articles

Bias in the Air: A Nationwide Exploration of Teachers’ Implicit Racial Attitudes, Aggregate Bias, and Student Outcomes

Theory suggests that teachers’ implicit racial attitudes affect their students, but large-scale evidence on U.S. teachers’ implicit biases and their correlates is lacking. Using nationwide data from Project Implicit, we found that teachers’ implicit White/Black biases (as measured by the implicit association test) vary by teacher gender and race. Teachers’ adjusted bias levels are lower in counties with larger shares of Black students. In the aggregate, counties in which teachers hold higher levels of implicit and explicit racial bias have larger adjusted White/Black test score inequalities and White/Black suspension disparities.

Impact of lean operations on the roles of finance functions and their application of lean

PurposeThis paper aims to study the relations between lean operations, lean principles in finance functions and the roles of finance functions.Design/methodology/approachThe paper uses structural equation modeling to analyze data from 408 different firms in the Danish production and services sectors. A dyadic approach is applied, as a sub-sample of 107 chief operating officers in the responding firms is used to investigate the construct validity, reliability and average deviation index of the instrument measuring the roles of finance functions.FindingsThe paper finds that lean-operation firms emphasize four different yet interdependent roles of finance functions. The paper also finds that lean operation leads to firms’ finance functions adopting lean principles.Research limitations/implicationsThis paper characterizes lean-operation firms as contextually ambidextrous to predict relations between lean operation and roles of finance functions. The paper expands prior case study findings on the roles of finance functions in lean-operation firms, and the findings of the paper underline that finance functions continue to play an important role in these firms.Practical implicationsDecision-makers in lean-operation firms should not be hesitant with respect to integrating finance function workers into the lean operation. Furthermore, decision-makers should understand that a balanced emphasis of the roles of finance functions is necessary to avoid overemphasizing exploitation at the expense of exploration, or vice versa.Originality/valueTo the best of the authors’ knowledge, this is the first paper to provide large-scale evidence of the roles of finance functions in lean-operation firms and to show that lean principles diffuse to finance functions. Furthermore, the paper introduces a new instrument for measuring finance function roles, based on the competing values framework.

Advanced services and differentiation advantage: an empirical investigation

PurposeThis study theoretically articulates and empirically validates a model of relationships between market complexity (competition intensity, heterogeneity and technological change), strategic focus on product and service differentiation, ADS offerings and differentiation advantage.Design/methodology/approachThe authors develop and test hypotheses through structural equation modeling based on data from the Sixth International Manufacturing Strategy Survey (IMSS-VI), involving 931 manufacturers from 22 countries.FindingsThe results indicate that (1) market complexity has a positive impact on strategic focus on product and service differentiation; (2) focus on product and service differentiation, but not market complexity, has a positive impact on the extent to which business units offer ADS to their customers; (3) ADS have a positive impact on service differentiation advantage, but no influence on product differentiation advantage.Practical implicationsManagers should incorporate decisions related to ADS provision as part of their manufacturing strategy formulation processes to align markets, strategic focus on product and service differentiation, and ADS provision. ADS seem an appropriate lever for market differentiation, because they appear not only to support service differentiation advantage, but also to be consistent with strategic focus on product differentiation.Originality/valueThe study provides novel insights and large-scale empirical evidence on the influence of the market environment on the offering of ADS, as well as on how relationships between the product and service activity in the manufacturing organization may affect differentiation advantage.

Long-Term Safety and Effectiveness of the Xanthine Oxidoreductase Inhibitor, Topiroxostat in Japanese Hyperuricemic Patients with or Without Gout: A 54-week Open-label, Multicenter, Post-marketing Observational Study.

Background and ObjectivesTopiroxostat, a selective xanthine oxidoreductase inhibitor, is used for the management of hyperuricemic patients with or without gout in Japan. Accumulating evidence has demonstrated the efficacy of topiroxostat for the treatment of hyperuricemia with or without gout. However, the safety and efficacy of topiroxostat in the clinical setting remain unclear, and there is little large-scale clinical evidence. We conducted a post-marketing observational study over 54 weeks.Patients and MethodsPatients were centrally enrolled, and case report forms of 4491 patients were collected between April 2014 and March 2019 from 825 medical sites.ResultsOverall, 4329 patients were assessed for safety and 4253 patients for effectiveness. The overall incidence of adverse drug reactions was 6.95%, and the incidence rates of adverse drug reactions of gouty arthritis, hepatic dysfunction, and skin disorders, which are of special interest in this study, were 0.79%, 1.73%, and 0.95%, respectively. No case of serious gouty arthritis was observed. Serum urate levels decreased stably over time and showed a significant reduction rate at 54 weeks (21.19% ± 22.07%) and on the final visit (19.91% ± 23.35%) compared to the baseline. The rates for subjects who achieved serum uric acid levels ≤ 6.0 mg/dL at 18 and 54 weeks after administration were 43.80% and 48.28%, respectively.ConclusionsThis study suggests that there is no particular concern about adverse drug reactions or the efficacy of topiroxostat for hyperuricemic patients with or without gout in a post-marketing setting in Japan.Electronic supplementary materialThe online version of this article (10.1007/s40261-020-00941-3) contains supplementary material, which is available to authorized users.

Fetal Programming of Semen Quality (FEPOS) Cohort - A DNBC Male-Offspring Cohort.

BackgroundPrenatal exposures may contribute to male infertility in adult life, but large-scale epidemiological evidence is still lacking. The Fetal Programming of Semen quality (FEPOS) cohort was founded to provide means to examine if fetal exposures can interfere with fetal reproductive development and ultimately lead to reduced semen quality and reproductive hormone imbalances in young adult men.MethodsYoung adult men at least 18 years and 9 months of age born to women in the Danish National Birth Cohort living in relative proximity to Copenhagen or Aarhus and for whom a maternal blood sample and two maternal interviews during pregnancy were available were invited to FEPOS. Recruitment began in March 2017 and ended in December 2019. The participants answered a comprehensive questionnaire and underwent a physical examination where they delivered a semen, urine, and hair sample, measured their own testicular volume, and had blood drawn.ResultsIn total 21,623 sons fulfilled eligibility criteria of whom 5697 were invited and 1058 participated making the response rate 19%. Semen characteristics did not differ between sons from the Copenhagen and Aarhus clinics. When comparing the FEPOS semen parameters to similar cohorts, the median across all semen characteristics was slightly lower for FEPOS participants, although with smaller variation.ConclusionWith its 1058 young adult men, the FEPOS cohort is the largest population-based male-offspring cohort worldwide specifically designed to investigate prenatal determinants of semen quality. Wide-ranging information on maternal health, lifestyle, socioeconomic status, occupation, and serum concentrations of potential reproductive toxicants during pregnancy combined with biological markers of fertility in their sons collected after puberty allow for in-depth investigations of the ‘fetal origins of adult disease hypothesis’.

Do Managers Learn from Institutional Investors Through Direct Interactions?

I examine whether corporate managers learn from institutional investors through direct interactions at investor conferences. While institutional investors often possess industry expertise, there is limited large-scale evidence on how they can provide useful information to corporate managers. I find that managers seek more direct interactions with institutional investors at conferences when they have a greater need for information about their firms’ product markets and supply chains. This relation is stronger when managers expect investors to be knowledgeable about these topics. I show that the information learned from conference interactions is reflected in managers’ subsequent decisions in that they issue forecasts more frequently and accurately. I also find that their personal stock trades during the seven days following conference participation earn higher abnormal returns than non-participating insider trades in the same firm-month. Overall, my paper shows that managers acquire decision-relevant information from their interactions with their firms’ institutional investors.

Utilising learning analytics to support study success in higher education: a systematic review

Study success includes the successful completion of a first degree in higher education to the largest extent, and the successful completion of individual learning tasks to the smallest extent. Factors affecting study success range from individual dispositions (e.g., motivation, prior academic performance) to characteristics of the educational environment (e.g., attendance, active learning, social embeddedness). Recent developments in learning analytics, which are a socio-technical data mining and analytic practice in educational contexts, show promise in enhancing study success in higher education, through the collection and analysis of data from learners, learning processes, and learning environments in order to provide meaningful feedback and scaffolds when needed. This research reports a systematic review focusing on empirical evidence, demonstrating how learning analytics have been successful in facilitating study success in continuation and completion of students’ university courses. Using standardised steps of conducting a systematic review, an initial set of 6220 articles was identified. The final sample includes 46 key publications. The findings obtained in this systematic review suggest that there are a considerable number of learning analytics approaches which utilise effective techniques in supporting study success and students at risk of dropping out. However, rigorous, large-scale evidence of the effectiveness of learning analytics in supporting study success is still lacking. The tested variables, algorithms, and methods collected in this systematic review can be used as a guide in helping researchers and educators to further improve the design and implementation of learning analytics systems.

Understanding the Complexity of Teacher Emotions From Online Forums: A Computational Text Analysis Approach.

Teacher emotions are complex as emotions are unique to individuals, situated within specific contexts, and vary over time. This study contributed in synthesizing theories of the complexity in two characteristics of multi-dimensionality and dynamics. Further, we provided large-scale empirical evidence by employing big data and computational text analysis. The data contained around one million teachers’ online posts from 2007 to 2018. It was scraped from three representative forums of teachers’ workplace events and personal life occasions in a popular American teacher website. By conducting thread-level sentiment analysis in forums, we computed word-frequency-based eight discrete emotions ratios (i.e., anger, anticipation, disgust, fear, joy, sadness, surprise, and trust) and the degrees of sentiment polarity (i.e., positive, negative, and neutral). We then used latent Dirichlet allocation for topic classifications. These topics, proxies of contexts, covered a holistic range of teachers’ real-life events. Some topics are in the main interest of scholars, such as teachers’ professional development and students’ behavioral management. This paper is also the first to include the less scholarly studied contexts like professional dressing advice and holiday choices. Then, we examined and visualized variations of emotions and sentiments across 30 topics along with three scales of time (i.e., calendar year, calendar month, and academic semesters). The results showed that teachers tended to have positive sentiments in the online professional community across the past decade, but all eight discrete emotions were presented. The compositions of the specific emotion types varied across topics and time. Regarding the topics of students’ behavior issues, teachers’ negative emotions’ ratios were higher compared when it was presented in other topics. Their negative emotions also peaked during semesters. The forum of teachers’ personal lives had positive emotions pronounced across topics and peaked during the wintertime. This paper summarized the evidenced multi-dimensionality characteristic with the multiple types of emotions as compositions and varying degrees of sentiment polarity of teachers. The dynamics characteristic is that teachers’ emotions vary across contexts from their workplace to their personal lives and over time. These two characteristics of complexity also suggested potential interplay effects among emotions and across contexts over time.

The Smart Home: How Consumers Craft New Service Networks by Combining Heterogeneous Smart Domestic Products

Service research suggests homes are becoming increasingly connected as consumers automate and personalize new forms of service provision. Yet, large-scale empirical evidence on how and why consumers automate smart domestic products (SDPs) is lacking. To address this knowledge gap, we analyze 13,905 consumer-crafted, automated combinations of SDPs, totaling 1,144,094 installations, across 253 separate service providers on the web service IFTTT.com. An exploratory network analysis examines the topology of the network and an interpretive coding exercise reveals how consumers craft different styles of human-computer interaction to cocreate value. The results reveal that the SDP network is disassortative, is imbalanced, has a long-tailed degree distribution, and that popular services have high centrality across all product category combinations. We show that popular combinations of SDPs are primarily motivated by utilitarian value-seeking enacted through a preference for automated tasks outside of conscious attention, though more individualistic combinations are slightly more likely to be hedonistically inclined. We conclude by showing how these consumer-crafted forms of service provision within domestic environments reveal design redundancy and opportunities for service innovation.

Combining the potential resilience of avian communities with climate change scenarios to identify areas of conservation concern

Abstract This study aimed to investigate the match between breeding bird communities’ potential resilience and projections of climate change in Europe. Here we identified European regions with the most substantial projected impacts of climate change based on Δ temperature and Δ precipitation in the next 60 years, assessing the overlap with maps of potential bird community resilience. We combined data on the number of species and functional redundancy of avian communities, to calculate an index of potential community resilience. Finally, combining these three layers of information, we obtained unique large-scale evidence of differences in potential conservation threats in the continent. Approximately 3% of the continent could be exposed to a maximum risk of conservation concern (areas characterized by more significant changes in precipitation and temperatures and simultaneously by avian communities with the lower functional redundancy) by 2070, with a 31% exposed to high risks, and 23% of the continent facing potentially moderate risk. Our findings provide important information on the potential capacity of European breeding bird communities to reduce the negative impact of changes in climate (temperature and precipitation), as well as identifying those regions potentially facing higher conservation risks (e.g. Southern part of Western Europe and the Ural Mountains in Russia).

C-reactive protein, interleukin-6 and pre-eclampsia: large-scale evidence from the GenPE case-control study

Multiple small studies have suggested that women with pre-eclampsia present elevated levels of C-reactive protein (CRP) and interleukin-6 (IL-6). However, little is known regarding the source of this CRP and IL-6 increase. Therefore, the aim of this study was to evaluate the relationship between CRP and IL-6 levels with pre-eclampsia considering different confounding factors. Using data from a large Colombian case-control study (3,590 cases of pre-eclampsia and 4,564 normotensive controls), CRP and IL-6 levels were measured in 914 cases and 1297 controls. The association between maternal serum levels of CRP and IL-6 with pre-eclampsia risk was evaluated using adjusted logistic regression models. Pre-eclampsia was defined as presence of blood pressure ≥140/90 mmHg and proteinuria ≥300mg/24 h (or ≥1 + dipstick). There was no evidence of association between high levels of CRP and IL-6 with pre-eclampsia after adjusting for the following factors: maternal and gestational age, ethnicity, place and year of recruitment, multiple-pregnancy, socio-economic position, smoking, and presence of infections during pregnancy. The adjusted OR for 1SD increase in log-CRP and log-IL-6 was 0.96 (95%CI 0.85, 1.08) and 1.09 (95%CI 0.97, 1.22), respectively. Although previous reports have suggested an association between high CRP and IL-6 levels with pre-eclampsia, sample size may lack the sufficient power to draw robust conclusions, and this association is likely to be explained by unaccounted biases. Our results, the largest case-control study reported up to date, demonstrate that there is not a causal association between elevated levels of CRP and IL-6 and the presence of pre-eclampsia.

Diagnostic and Prognostic Value of Ergonovine Echocardiography for Noninvasive Diagnosis of Coronary Vasospasm

ObjectivesThis study sought to obtain large-scale evidence supporting the clinical usefulness of ergonovine echocardiography. BackgroundThe role of noninvasive ergonovine provocation testing with echocardiographic monitoring of ventricular wall motion (ergonovine echocardiography) needs to be defined. MethodsClinical data of patients who underwent ergonovine echocardiography in 3 tertiary referral hospitals in South Korea were analyzed. ResultsErgonovine echocardiography was performed in 14,012 patients (mean age 52.8 ± 11.1 years; 6,213 [44.3%] women) after exclusion of significant coronary arterial stenosis by functional (treadmill or perfusion scan, n = 9,824) or anatomic test (invasive or computerized tomographic coronary angiography, n = 4,188). Premature termination developed in 0.4% (n = 51), and a positive result was observed in 2,144 patients (15.3%), with variable frequencies according to the diagnosis (acute coronary syndrome [38.2%], variant angina [31.8%], effort angina [14.9%], aborted sudden cardiac death [17.6%], syncope [9.9%]). There was no mortality or development of myocardial infarction during the test. During median follow-up of 11.4 (interquartile range: 7.2 to 15.8) years, death of any cause and cardiovascular death occurred in 494 and 143 patients, respectively. The 10-year overall (96.7 ± 0.2% vs. 91.5 ± 0.6%; p < 0.0001) and cardiovascular mortality–free (99.2 ± 0.1% vs. 96.7 ± 0.4%; p < 0.0001) survival rates were lower in patients with positive ergonovine echocardiography. Regarding patients with positive test results, the functional test group and the anatomic test group did not show a significant difference in the survival rates. After adjustment of age and male sex, a positive test was an independent risk factor associated with all-cause mortality (hazard ratio: 1.879, 95% confidence interval: 1.548 to 2.280; p < 0.001) and cardiovascular death (hazard ratio: 2.903, 95% confidence interval: 2.061 to 4.089; p < 0.001). ConclusionsErgonovine echocardiography for coronary vasospasm diagnosis could be safely performed even without angiographic documentation of fixed coronary stenosis depending on the clinical presentation, and provided important prognostic implication. Ergonovine echocardiography can replace the invasive spasm provocation testing, which has been overlooked unfairly.

The Bright Side of Tax Avoidance: The Case of Foreign Competition

We investigate the link between foreign competition and corporate tax planning. We find that a one standard deviation increase in import competition leads to a reduction of 9% in firms’ effective tax rates relative to standard levels. Further, we document that our evidence is robust to a battery of sensitivity tests and endogeneity concerns. Cross-sectional analysis shows that the association between foreign competition and tax planning is most pronounced among firms that: (i) are financially constrained; (ii) have the greatest ability to implement tax avoidance policies; (iii) have strong alignment of incentives between managers and shareholders; (iv) appoint high-quality auditors; and (v) attract more analyst coverage. Additional tests reveal that tax planning in the presence of foreign competitive threats enhances firm performance and investments, and improves managers’ career outcomes. Overall, we provide the first large-scale evidence that changes in the macroeconomic external environment are an important determinant of corporate tax planning activities.

Lost or Found? Discovering Data Needed for Research

Finding data is a necessary precursor to being able to reuse data, although relatively little large-scale empirical evidence exists about how researchers discover, make sense of and (re)use data for research. This study presents evidence from the largest known survey investigating how researchers discover and use data that they do not create themselves. We examine the data needs and discovery strategies of respondents, propose a typology for data reuse and probe the role of social interactions and literature search in data discovery. We consider how data communities can be conceptualized according to data uses and propose practical applications of our findings for designers of data discovery systems and repositories. Specifically, we consider how to design for a diversity of practices, how communities of use can serve as an entry point for design and the role of metadata in supporting both sensemaking and social interactions.

The Invisible Risk: Pandemics and the Financial Markets

Are pandemics systemically important to modern-day financial markets? This study uses the COVID-19 pandemic as a natural experiment for testing how large-scale pandemics affect the financial markets. Using hand-collected data at the firm level, I find that managers systematically underestimated their exposure to pandemics in their SEC-mandated risk factors, and the vast majority of firms decreased in value at the pandemic's onset. I also find that the pandemic triggered unprecedented changes in U.S. employment levels and the values of bonds, commodities, and currencies. These types of findings suggest that pandemics are systemically important to the financial markets. Overall, this study provides some of the first large-scale evidence on how pandemics affect the financial markets.

Daily use of high-potency cannabis is associated with more positive symptoms in first-episode psychosis patients: the EU-GEI case-control study.

Daily use of high-potency cannabis has been reported to carry a high risk for developing a psychotic disorder. However, the evidence is mixed on whether any pattern of cannabis use is associated with a particular symptomatology in first-episode psychosis (FEP) patients. We analysed data from 901 FEP patients and 1235 controls recruited across six countries, as part of the European Network of National Schizophrenia Networks Studying Gene-Environment Interactions (EU-GEI) study. We used item response modelling to estimate two bifactor models, which included general and specific dimensions of psychotic symptoms in patients and psychotic experiences in controls. The associations between these dimensions and cannabis use were evaluated using linear mixed-effects models analyses. In patients, there was a linear relationship between the positive symptom dimension and the extent of lifetime exposure to cannabis, with daily users of high-potency cannabis having the highest score (B = 0.35; 95% CI 0.14-0.56). Moreover, negative symptoms were more common among patients who never used cannabis compared with those with any pattern of use (B = -0.22; 95% CI -0.37 to -0.07). In controls, psychotic experiences were associated with current use of cannabis but not with the extent of lifetime use. Neither patients nor controls presented differences in depressive dimension related to cannabis use. Our findings provide the first large-scale evidence that FEP patients with a history of daily use of high-potency cannabis present with more positive and less negative symptoms, compared with those who never used cannabis or used low-potency types.

Comparison of Cardiovascular and Safety Outcomes of Chlorthalidone vs Hydrochlorothiazide to Treat Hypertension

Chlorthalidone is currently recommended as the preferred thiazide diuretic to treat hypertension, but no trials have directly compared risks and benefits. To compare the effectiveness and safety of chlorthalidone and hydrochlorothiazide as first-line therapies for hypertension in real-world practice. This is a Large-Scale Evidence Generation and Evaluation in a Network of Databases (LEGEND) observational comparative cohort study with large-scale propensity score stratification and negative-control and synthetic positive-control calibration on databases spanning January 2001 through December 2018. Outpatient and inpatient care episodes of first-time users of antihypertensive monotherapy in the United States based on 2 administrative claims databases and 1 collection of electronic health records were analyzed. Analysis began June 2018. Chlorthalidone and hydrochlorothiazide. The primary outcomes were acute myocardial infarction, hospitalization for heart failure, ischemic or hemorrhagic stroke, and a composite cardiovascular disease outcome including the first 3 outcomes and sudden cardiac death. Fifty-one safety outcomes were measured. Of 730 225 individuals (mean [SD] age, 51.5 [13.3] years; 450 100 women [61.6%]), 36 918 were dispensed or prescribed chlorthalidone and had 149 composite outcome events, and 693 337 were dispensed or prescribed hydrochlorothiazide and had 3089 composite outcome events. No significant difference was found in the associated risk of myocardial infarction, hospitalized heart failure, or stroke, with a calibrated hazard ratio for the composite cardiovascular outcome of 1.00 for chlorthalidone compared with hydrochlorothiazide (95% CI, 0.85-1.17). Chlorthalidone was associated with a significantly higher risk of hypokalemia (hazard ratio [HR], 2.72; 95% CI, 2.38-3.12), hyponatremia (HR, 1.31; 95% CI, 1.16-1.47), acute renal failure (HR, 1.37; 95% CI, 1.15-1.63), chronic kidney disease (HR, 1.24; 95% CI, 1.09-1.42), and type 2 diabetes mellitus (HR, 1.21; 95% CI, 1.12-1.30). Chlorthalidone was associated with a significantly lower risk of diagnosed abnormal weight gain (HR, 0.73; 95% CI, 0.61-0.86). This study found that chlorthalidone use was not associated with significant cardiovascular benefits when compared with hydrochlorothiazide, while its use was associated with greater risk of renal and electrolyte abnormalities. These findings do not support current recommendations to prefer chlorthalidone vs hydrochlorothiazide for hypertension treatment in first-time users was found. We used advanced methods, sensitivity analyses, and diagnostics, but given the possibility of residual confounding and the limited length of observation periods, further study is warranted.

Transgender Status, Gender Identity, and Socioeconomic Outcomes in the United States

This article provides the first large-scale evidence on transgender status, gender identity, and socioeconomic outcomes in the United States, using representative data from 35 states in the Behavioral Risk Factor Surveillance System (BRFSS), which asked identical questions about transgender status and gender identity during at least one year from 2014 to 2017. More than 2,100 respondents, aged 18 to 64 years, identified as transgender. Individuals who identify as transgender are significantly less likely to be college educated and less likely to identify as heterosexual than are individuals who do not identify as transgender. Controlling for these and other observed characteristics, transgender individuals have significantly lower employment rates, lower household incomes, higher poverty rates, and worse self-rated health compared to otherwise similar men who are not transgender.

Is the Social Safety Net a Long-Term Investment? Large-Scale Evidence from the Food Stamps Program

We use novel, large-scale data on 43 million Americans from the 2000 Census and the 2001 to 2013 American Communities Survey linked to the Social Security Administration’s NUMIDENT to study how a policy-driven increase in economic resources for families affects children’s long-term outcomes. Using variation from the county-level roll-out of the Food Stamps program between 1961 and 1975, we find that children with access to greater economic resources before age five experience an increase of 6 percent of a standard deviation in their adult human capital, 3 percent of a standard deviation in their adult economic self-sufficiency, 8 percent of a standard deviation in the quality of their adult neighborhoods, 0.4 percentage-point increase in longevity, and a 0.5 percentage-point decrease in likelihood of being incarcerated. Based on these estimates, we conclude that Food Stamps’ transfer of resources to families is a highly cost-effective investment into young children, yielding a marginal value of public funds of approximately 56. Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.

Robots and the Gender Pay Gap in Europe

Could robotization make the gender pay gap worse? We provide the first large-scale evidence on the impact of industrial robots on the gender pay gap using data from 20 European countries. We show that robot adoption increases both male and female earnings but also increases the gender pay gap. Using an instrumental variable strategy, we find that a ten percent increase in robotization leads to a 1.8 percent increase in the gender pay gap. These results are mainly driven by countries with high levels of gender inequality and outsourcing destination countries. We then explore the mechanisms behind this effect and find that our results can be explained by the fact that men at medium- and high-skill occupations disproportionately benefit from robotization (through a productivity effect). We rule out the possibility that our results are driven by mechanical changes in the gender composition of the workforce nor by inflows or outflows from the manufacturing sector.