Machine Learning and Perceived Age Stereotypes in Job Ads: Evidence from an Experiment

Ian Burn, Daniel Firoozi, Daniel Ladd, David Neumark | January, 2021

We explore whether ageist stereotypes in job ads are detectable using machine learning methods measuring the linguistic similarity of job-ad language to ageist stereotypes identified by industrial psychologists. We then conduct an experiment to evaluate whether this language is perceived as biased against older workers. We find that language classified by the machine learning algorithm as closely related to ageist stereotypes is perceived as ageist by experimental subjects. The scores assigned to the language related to ageist stereotypes are larger when responses are incentivized by rewarding participants for guessing how other respondents rated the language. These methods could potentially help enforce anti-discrimination laws by using job ads to predict or identify employers more likely to be engaging in age discrimination.

 Download PDF  


The Career Evolution of the Sex Gap in Wages: Discrimination vs. Human Capital Investment

David Neumark, Giannina Vaccaro | December, 2020

Several studies find that there is little sex gap in wages at labor market entry, and that the sex gap in wages emerges (and grows) with time in the labor market. This evidence is consistent with (i) there is little or no sex discrimination in wages at labor market entry, and (ii) the emergence of the sex gap in wages with time in the labor market reflects differences between men and women in human capital investment (and other decisions), with women investing less early in their careers. Indeed, some economists explicitly interpret the evidence this way. We show that this interpretation ignores two fundamental implications of the human capital model, and that differences in investment can complicate the interpretation of both the starting sex gap in wages (or absence of a gap), and the differences in “returns” to experience. We then estimate stylized structural models of human capital investment and wage growth to identify the effects of discrimination and differences in human capital investment, and find evidence more consistent with discrimination reducing women’s wages at labor market entry. 

 Download PDF  


Happy Together? The Peer Effects of Dual Enrollment Students on Community College Student Outcomes

Vivian Yuen Ting Liu, Di Xu | November, 2020

Nationally, 15% of first-time community college students were high school dual enrollment (DE) students, which raises concerns about how high school peers might influence college enrollees. Using administrative data from a large state community college system, we examine whether being exposed to a higher percentage of DE peers in entry-level (gateway) math and English courses influences non-DE enrollees’ performance. Using a two-way fixed effects model, our results indicate that college enrollees exposed to a higher proportion of DE peers had lower pass rates and grades in gateway courses, and higher course repetition rates. Supplemental student-level analysis suggests that greater exposure to DE peers during a student’s initial semester in college reduces next-term college persistence.

 Download PDF


Initial Host-Society/Migrant Relations: Implications for U.S. Refugee Integration
Thoa V. Khuu, Frank D. Bean | October, 2020

Research into factors affecting immigrant integration carries important implications for immigration scholars and policymakers. By immigrant integration we mean the nature and extent of temporal and generational convergence between newcomers and natives in sociocultural patterns and socioeconomic attainment (Brown and Bean 2006; Jimenez 2016). Although many studies have investigated the extent to which immigrants and natives come to resemble one another (Waters and Pineau 2015), fewer have devoted attention to whether newcomers arriving under various entry auspices exhibit different integration dynamics and outcomes. A notable exception involves research assessing the degree to which unauthorized entrants incur substantial handicaps compared to those entering with legal status. Because the United States has largely failed to extend official societal membership to unauthorized migrants, their families have been deprived of access to opportunities for achieving socioeconomic mobility (Brown and Bean 2016). Research shows that this has negatively affected migrants, their migrating children, and even their children born in the United States (e.g., Bean, Brown and Bachmeier 2015; Gonzales 2015). Although numerous studies provide striking examples of how this kind of host-society/migrant relationship strongly affects migrant integration, little investigation has delved into the nature and degree to which immigrants arriving under alternative forms of legal entry undergo different integration experiences. 

 Download PDF  


The Causes and Consequences of the 1986 Immigration Reform and Control Act (IRCA)
Frank D. Bean, Thoa V. Khuu | October, 2020

The United States often views itself as a nation of immigrants. Because of this, in part since the beginning of the twentieth century, it has only rarely adopted major changes in its immigration policy.  Until the reforms of 1986, only the 1924 National Origins Quota Act and its modification in 1965 (through amendments to the 1952 McCarran Walter Act) involved substantial reform.  This changed with the passage of the 1986 Immigration Reform and Control Act (IRCA) and its derivative sequel, the 1990 Immigration Act.  And as of this writing in 2020, no other substantial pieces of immigration legislation have been passed by Congress.  IRCA emerged from and followed in considerable measure the recommendations of the Select Commission on Immigration and Refugee Policy (1979-1981). That body sought to reconcile two competing political constituencies, one favoring greater immigration restriction and the other an expansion of family-based and work-related migration. The IRCA legislation contained something for each side: the passage of employer sanctions, or serious penalties on employers for hiring unauthorized workers, for the restriction side; and the provision of a legalization program, which outlined a pathway for certain unauthorized entrants to obtain green cards and eventually citizenship, for the reform side. The complete legislative package also included other provisions: including criteria for the admission of  agricultural workers, a measure providing financial assistance to states for the costs they would incur from migrants legalizing, a requirement that states develop ways to verify that migrants were eligible for welfare benefits, and a provision providing substantial boosts in funding for border enforcement activities. In the years after the enactment of IRCA, research has revealed that the two major compromise provisions plus the agricultural provision have generated mixed results. Employer sanctions failed to curtail unauthorized migration much, in all likelihood because of minimal funding for enforcement, while legalization and the agricultural measure resulted in widespread enrollment, with almost all of the unauthorized migrants who qualified for it coming forward to take advantage of the opportunity to become U.S. legalized permanent residents (LPRs). In general, however, IRCA can be interpreted in political historical terms as exemplifying contradictory parts. On the one hand, its somewhat expansionist and its legalization elements reflect the inclusive/pluralistic tendencies of much of 18th century immigration, but its employer sanction provisions contained the seeds for the subsequent development of restrictive/exclusive socio-political tendencies.

 Download PDF 


Finding and Keeping Friends in College and their Influence on Alcohol Use: A Network Analysis
David R. Schaefer, Irene van Woerden, Daniel Hruschka, Meg Bruening | September, 2020

Objective. We investigate how alcohol use and friendship co-evolve during students’ transition to university. We discern effects of peer influence from friend selection based on alcohol use, whether such effects vary in strength across the school year, and whether alcohol has different effects on friendship formation versus friendship maintenance. Method. We gathered data on friendships, alcohol use, and binge drinking from 300 residence hall students (71% female) at a large, public U.S. university. Surveys were conducted at four time points during the 2015-16 academic year. We used a stochastic actor-oriented model (SAOM) to test whether alcohol use was influenced by one’s friends, while simultaneously testing for friend selection based on alcohol use and related network processes. Results. Students were 7.0 times more likely to drink alcohol weekly if all vs. none of their friends drank weekly, and 6.8 times more likely to binge drink when all vs. none of their friends engaged in binge drinking, after controlling for friend selection. Alcohol use differentially affected friendship creation and maintenance in a complex manner (1) weekly drinkers were more likely to form new friendships and dissolve existing friendships than non-drinkers; and (2) similarity on drinking fostered new friendships, but had no effect on friendship persistence. Conclusions. Friends influence one another’s weekly drinking and binge drinking, while conversely, alcohol use contributes to both friendship formation and friendship instability.

 Download PDF 


Modifiable Kindergarten Factors that Predict Being a Bully, Victim, or Bully-victim by the Upper Elementary Grades
Paul L. Morgan, Adrienne D. Woods, Yangyang Wang, George Farkas, Marianne M. Hillemeier, Yoonkyung Oh | September, 2020

The investigators analyzed a population-based cohort (N range=7,182-8,210; kindergarten Mage =67.5 months) to identify modifiable factors by kindergarten predictive of being a bully, victim, or bully-victim during third, fourth, or fifth grade. Chi-squared analyses supported the bully-victim subtype. Greater academic achievement lowered children’s risk for being bullies (odds ratio [OR] range = .66 to .75), victims (OR =.83 to .85), and bully-victims (OR = .72 to .76). Externalizing problem behaviors increased children’s risk for being bullies (OR = 1.84 to 2.16), victims (OR = 1.35 to 1.43), and bully-victims (OR = 1.90 to 2.17). Internalizing problem behaviors and parenting did not consistently predict children’s bullying victimization. Achievement and behavior but not parenting constitute modifiable targets of early bullying victimization interventions.

 Download PDF 


Disease, Downturns, and Wellbeing: Economic History and the Long-Run Impacts of COVID-19
Vellore Arthi, John Parman | September, 2020

How might COVID-19 affect human capital and wellbeing in the long run? The COVID-19 pandemic has already imposed a heavy human cost—taken together, this public health crisis and its attendant economic downturn appear poised to dwarf the scope, scale, and disruptiveness of most modern pandemics. What evidence we do have about other modern pandemics is largely limited to short-run impacts. Consequently, recent experience can do little to help us anticipate and respond to COVID-19’s potential long-run impact on individuals over decades and even generations. History, however, offers a solution. Historical crises offer closer analogues to COVID-19 in each of its key dimensions—as a global pandemic, as a global recession—and offer the runway necessary to study the life-course and intergenerational outcomes. In this paper, we review the evidence on the long-run effects on health, labor, and human capital of both historical pandemics (with a focus on the 1918 Influenza Pandemic) and historical recessions (with a focus on the Great Depression). We conclude by discussing how past crises can inform our approach to COVID-19—helping tell us what to look for, what to prepare for, and what data we ought to collect now.

 Download PDF 


How Valuable are Civil Liberties? Evidence from Gang Injunctions and Housing Prices in Southern California
Emily Owens, Michelle D. Mioduszewski, Christopher J. Bates | August, 2020

Place-based and proactive policing strategies can reduce crime. However, the broader net impacts of these policies on targeted communities has yet to be quantified, meaning there is little empirical evidence on if, or when, policing is socially beneficial. Using a spatial discontinuity in constraints on police actions created by civil gang injunctions and temporal variation in when injunctions are enacted, we find that aggressive policing can reduce, rather than increase, people’s desire to live in affected neighborhoods. Mover demographics suggest that homebuyers perceive injunction areas as safe places, but where negative police encounters are common. Dividing our sample by pre-injunction crime rates suggest that net willingness-to-pay to avoid aggressive police encounters falls as the possible expected benefit from crime reduction increases.

 Download PDF


Understanding Socioeconomic Disparities in Travel Behavior during the COVID-19 Pandemic
Rebecca Brough, Matthew Freedman, and David C. Phillips | June, 2020

We document the magnitudes of and mechanisms behind socioeconomic differences in travel behavior during the COVID-19 pandemic. We focus on King County, Washington, one of the rst places in the U.S. where COVID-19 was detected. We leverage novel and rich administrative and survey data on travel volumes, modes, and preferences for different demographic groups. Large average declines in travel, and in public transit use in particular, due to the pandemic and related policy responses mask substantial heterogeneity across socioeconomic groups. Travel intensity declined considerably less among less-educated and lower-income individuals, even after accounting for mode substitution and variation across neighborhoods in the impacts of public transit service reductions. The relative inability of less-educated and lower-income individuals to cease commuting explains at least half of the difference in travel responses across groups.

Download PDF


Older Workers Need Not Apply? Ageist Language in Job Ads and Age Discrimination in Hiring
Ian Burn, Patrick Button, Luis Munguia Corella, and David Neumark | May, 2020

We study the relationships between ageist stereotypes – as reflected in the language used in job ads – and age discrimination in hiring, exploiting the text of job ads and differences in callbacks to older and younger job applicants from a resume (correspondence study) field experiment (Neumark, Burn, and Button, 2019). Our analysis uses methods from computational linguistics and machine learning to directly identify, in a field-experiment setting, ageist stereotypes that underlie age discrimination in hiring. The methods we develop provide a framework for applied researchers analyzing textual data, highlighting the usefulness of various computer science techniques for empirical economics research. We find evidence that language related to stereotypes of older workers sometimes predicts discrimination against older workers. For men, our evidence points to age stereotypes about all three categories we consider – health, personality, and skill – predicting age discrimination, and for women, age stereotypes about personality. In general, the evidence is much stronger for men, and our results for men are quite consistent with the industrial psychology literature on age stereotypes.

Download PDF



© UC Irvine School of Social Sciences - 3151 Social Sciences Plaza, Irvine, CA 92697-5100 - 949.824.2766