ArticlePDF Available

Employment discrimination against older workers: An experimental study of hiring practices



Pairs of resumes, one for a 57-year-old and the other for a 32-year-old, were mailed to 775 large firms and employment agencies across the United States. Although the resumes presented equal qualifications, the older job seeker received a less favorable employer response 26.5% of times when a position appeared to be vacant. Vigorous enforcement of equal opportunity laws as well as initiatives to change employer attitudes are appropriate responses to such discrimination. The technique of employment testing, demonstrated in this research, can be useful in both efforts.
... Bendick, Jackson and Romero (1997) and Riach and Rich (2007) use correspondence testing in the US and the UK, respectively, both detected a significant negative age bias against older applicants. Bendick, Brown and Wall (1999) applied an audit study, in which pairs of testers of different ages applied by telephone for different occupational positions. ...
... Scholars had different solutions to that problem. Bendick et al. (1997) matched pairs of applicants aged 32 and 57 years for white-collar jobs. ...
... In addition, the older worker was equipped with irrelevant work experience for the current position to compensate the gap in work experience across life time. Overall, older applicants received a significantly higher number of negative responses to vacancies Contrary to Bendick et al. (1997), who adjusted fictitious applicants to the method, Riach and Rich (2007) claimed instead for adapting the technique to older workers. By doing that, the authors created matched pairs of applicants that differ in their job experience, since experience serves as an indicator for age. ...
The thesis investigates age discrimination in personnel selection decisions across different industries and positions in Austria. A factorial sur-vey approach (FSA), also called vignette study, is applied. This semi-experimental research method combines the advantages of both survey re-search and classical experiments. A vignette is a short description of a social situation or a person that consists of several characteristics (factors and their levels) which can be simultaneously manipulated. The experimental design takes on all possible combinations of factor levels across all factors. Such de-sign allows studying the effect of each factor on the dependent variable, as well as the effects of interactions between factors on the response variable. The impact of each dimension on respondent’s judgment can be estimated accurately. To make it more specific, through a vignette design it is possible to investigate the solely effect of age on selection decisions, a factor that is usually confounded with other factors in reality. Respondents in this study are HR managers from different organizations in Austria. They are confront-ed with scenarios of hypothetical applicants and are then asked to state the hiring desirability for the respective job seeker. The paper elaborates on the research method and presents results of multilevel analysis. Keywords: Age discrimination; personnel selection; Vignette approach
... Les années « manquantes » sont alors laissées à l'appréciation de l'employeur, en ne renseignant que les postes occupés les plus récents (Lahey, 2008b). Une autre option est de justifier cette absence de la manière la plus neutre possible, par exemple en indiquant pour les femmes que cette période a été consacrée à élever leurs enfants et pour les hommes à enseigner (Bendick et al., 1997). Le problème est que les signaux envoyés par ces différents types d'expérience ou non-expérience peuvent biaiser l'étude. ...
... La plupart des études ont été réalisées aux États-Unis (Bendick et al., 1997(Bendick et al., , 1999Neumark et al., 2015 ;Lahey, 2008). La seule étude publiée en France trouve que la discrimination sur l'âge y est particulièrement forte notamment en comparaison avec l'Allemagne et l'Angleterre (Riach et Rich, 2006a ;Riach, 2015). ...
... Depuis les lettres envoyées en candidature spontanée, ou en réponse à des petites annonces publiées dans les journaux, jusqu'à l'automatisation d'envoi de mails en réponse à des offres sur internet. Cette évolution et cette numérisation des processus de recrutement ont permis de grandement augmenter les échantillons considérés : de 1 550 CV envoyés pour le premier test de discrimination à l'embauche ayant considéré le motif de l'âge (Bendick et al., 1997) à 40 236 CV pour le plus récent (Neumark et al., 2015). ...
... Examples of such screening problems are: finding patients in a large database of electronic health records to manually evaluate for qualification to take part in a medical trial [1]; the first stage of a multi-stage retrieval pipeline of a search engine [2]; or which people to reach out to with a personalized invitation to apply to a specific job posting [3]. In each of these examples, there is significant pressure to make high-quality, unbiased screening decisions quickly and efficiently, often about thousands or even millions of candidates under limited resources and additional diversity requirements [4,5,6,2]. While these screening decisions have been made manually or through manually constructed rules in the past, automated predictive tools for optimizing screening decisions are becoming more prevalent [7,8,3,9]. ...
Many selection processes such as finding patients qualifying for a medical trial or retrieval pipelines in search engines consist of multiple stages, where an initial screening stage focuses the resources on shortlisting the most promising candidates. In this paper, we investigate what guarantees a screening classifier can provide, independently of whether it is constructed manually or trained. We find that current solutions do not enjoy distribution-free theoretical guarantees -- we show that, in general, even for a perfectly calibrated classifier, there always exist specific pools of candidates for which its shortlist is suboptimal. Then, we develop a distribution-free screening algorithm -- called Calibrated Subset Selection (CSS) -- that, given any classifier and some amount of calibration data, finds near-optimal shortlists of candidates that contain a desired number of qualified candidates in expectation. Moreover, we show that a variant of our algorithm that calibrates a given classifier multiple times across specific groups can create shortlists with provable diversity guarantees. Experiments on US Census survey data validate our theoretical results and show that the shortlists provided by our algorithm are superior to those provided by several competitive baselines.
... See:Bendick, Jackson, and Romero (1997);Bendick, Brown, and Wall (1999);Riach and Rich (2006, 2010); Lahey (2008); Baert et al. (2016); Farber, Silverman, and von Wachter (2017); Farber et al. (2019); Carlsson and Eriksson (2019); Neumark, Burn, and Button (2016, 2019); and Neumark et al. (2019).3 NBB provide an extensive discussion regarding the interpretation of resume-correspondence study findings as reflecting age discrimination. ...
... However, this is accompanied by many ethical concerns as we know from decades of research. Some employers have tended to discriminate against women and ethnic minorities, albeit unconsciously [7,8,9,10]. Contrary to popular belief, discrimination occurs not only in traditional recruiting, but also in recruiting algorithms. ...
Conference Paper
Full-text available
To reduce the workload of employees working in Human Resource departments and to avoid bias in pre-selection of applicants, an increasing number of companies deploy Artificial Intelligence (AI)-based algorithms. Some examples such as Amazon's discriminating recruiting algorithm showed that algorithms are not free of unethical decision making. Although there already exists a variety of ethics principles for AI-based systems, those are usually hardly being applicable to specific use cases such as using AI-based algorithms in recruiting processes. To address this issue and to provide guidance for researchers and practitioners, we conducted a systematic literature review (keyword and backwards search) on existing ethics guidelines and principles for AI and extracted aspects that seemed applicable to guide recruiting processed. Based on 28 relevant papers we derived actionable guidelines for using AI-based algorithms in recruiting processes. We categorized our guidelines into the aspects of fairness, avoidance of discrimination and avoidance of bias.
... Our research does not directly assess how individuals can mitigate forms of bias in the job application process. Existing literature suggests some success may come from highlighting attributes that contrast with the negative stereotype (Bendick, Jackson, & Romero, 1996); however, future research is warranted. Future research should consider other forms of employment discrimination against GM people as well as personal and organizational strategies for mitigating such discrimination. ...
Full-text available
We sought to examine perceived gender identity, perceived co-worker discomfort, and salary recommendations for youth counselors with transgender-related work experience. In two experiments conducted in 2016 and 2017, we randomized participants to view 1 of 2 résumés with varying work experience at a camp for transgender youth or a generic youth camp. Study 1 participants were 274 adult festivalgoers at a lesbian, gay, bisexual, and transgender pride festival. Study 2 participants were 296 employed, heterosexual adults aged 35-60 from an online survey panel. In Study 1, viewing the résumé with transgender experience resulted in a statistically significantly higher likelihood of reporting the applicant was gender minority than cisgender (adjusted odds ratio = 3.76, 95% confidence interval [1.32, 10.72], p = .01), higher but not a statistically significant level of co-worker discomfort (aOR = 1.39, 95% CI [0.83, 2.32], p = .22), and, although not statistically significant, a $2,605 higher salary (95% CI [-$604, - $5,814], p = .11). In Study 2, we found a statistically significantly greater likelihood of reporting the applicant was gender minority than cisgender (OR = 2.56, 95% CI [1.36, 4.82], p < .01), statistically significantly higher odds of reported co-worker discomfort (OR = 3.57, 95% CI [2.15, 5.92], p < .01), and, although not statistically significant, a $1,374 higher salary (95% CI [-$1,931, $4,679], p = .41). Our results indicate the potential for stigma by association for professionals working with marginalized groups and suggest potential pathways through which employment discrimination may exacerbate existing inequities for gender minority people.
... A well-known study by Bertrand and Mullainathan demonstrated that given identical resumes, employer response rates were significantly higher when candidates had names that suggested they were white males as compared to other groups [9]. This results of this study, and others like them [8,7,43], are widely accepted in the world of human resources, where practitioners continually seek new ways of handling bias [49]. Advocates argue for algorithmic techniques as a means to address bias [16,27], and indeed, there is some evidence that they can be used to combat human idiosyncrasies [40,46]. ...
There has been rapidly growing interest in the use of algorithms for employment assessment, especially as a means to address or mitigate bias in hiring. Yet, to date, little is known about how these methods are being used in practice. How are algorithmic assessments built, validated, and examined for bias? In this work, we document and assess the claims and practices of companies offering algorithms for employment assessment, using a methodology that can be applied to evaluate similar applications and issues of bias in other domains. In particular, we identify vendors of algorithmic pre-employment assessments (i.e., algorithms to screen candidates), document what they have disclosed about their development and validation procedures, and evaluate their techniques for detecting and mitigating bias. We find that companies' formulation of "bias" varies, as do their approaches to dealing with it. We also discuss the various choices vendors make regarding data collection and prediction targets, in light of the risks and trade-offs that these choices pose. We consider the implications of these choices and we raise a number of technical and legal considerations.
Full-text available
Entre as muitas mudanças e transformações em curso está o franco envelhecimento da população mundial. O Brasil, por sua vez, se deparará com o mesmo problema, já que as projeções demográficas indicam que haverá a aceleração dessa tendência em nosso país a partir de 2030. Dada a importância do assunto, é fundamental que políticas sociais preemptivas sejam colocadas em ação – algo, aliás, com o qual não estamos culturalmente acostumados. Mas é forçoso reconhecer que uma mudança de tal magnitude vem acompanhada de outros efeitos como, por exemplo, o envelhecimento da força de trabalho. Assim sendo, a minha intenção inicial ao escrever essa obra era de pura e simplesmente desvendar o universo do trabalhador mais velho (o talento maduro) e conhecer a sua realidade, já que ele deverá ocupar um papel-chave doravante. Afinal de contas, os avanços da medicina e o consequente aumento da longevidade favorecem a obtenção de carreiras profissionais mais longevas. Entretanto, cumpre destacar que a literatura mundial sobre o tema revela a existência de mitos e pressupostos infundados que, no geral, tendem a tornar a vida desse grupo particularmente difícil. E tal constatação não deixa de ser um já que estamos diante de um fenômeno social e demográfico praticamente generalizado, bem como o fato de que se trata de algo inerente às nossas existências. Em suma, pode-se afirmar que a velhice não é bem recebida. De alguma forma, ela evoca lembranças ou sinais indesejáveis – e o mesmo raciocínio vale para a esfera do trabalho. Como seres ainda imperfeitos, nós somos moldados por preconceitos e percepções estereotipadas decorrentes dos valores que abrigamos em nossos selfs. Embora possamos (e devemos) controlá-las, elas estão lá em algum lugar da nossa personalidade. Os ambientes de trabalho não estão imunes aos efeitos deletérios de tais visões e opiniões pessoais. Do contrário, nós viveríamos num mundo absoluta-mente perfeito – condição essa ainda inconquistada, como mostra a realidade. Mas ao escrutinizar a situação do trabalhador mais velho acabei, por exten-são, esbarrando numa problemática tão ou mais complexa, ou seja, o imperativo de promover a diversidade em nossas empresas e de desenvolver ambientes mais inclusivos. De fato, os primeiros estudos sobre diversidade no Brasil datam dos anos 1990. No entanto, ainda persistem muitas perguntas a ser respondidas e, sobre-tudo, ações a ser implementadas no âmbito das organizações. Dito de outra forma, os assuntos tratados nesse livro ainda não foram adequadamente assimilados. Dessa forma, portanto, os meus objetivos ao escrever essa obra foram consideráveis – ou seja: analisar criticamente três relevantes aspectos referentes à realidade dos ambientes de trabalho da atualidade. Ao encetar tal empreendimento, procurei deixar de lado as minhas próprias impressões e convicções construídas aprioristicamente e fundamentar as minhas conclusões em evidências empíricas (e aí reside a maior contribuição dessa obra). Por outro lado, escrever para públicos-alvos distintos é sempre muito desa-fiador dado que as necessidades são singulares. Nesse sentido, na medida do pos-sível busquei abastecer os praticantes de gestão da diversidade e líderes organiza-cionais com informações criteriosamente embasadas, ricos insightse conclusões fundamentadas, livrando-os da superficialidade observável em muitas obras do gênero. Por isso, os leitores pertencentes a esse grupo encontrarão subsídios para alicerçar suas decisões e projetos, assim como aspectos para confrontar e ponderar. Não há aqui, a rigor, uma receita adrede preparada. Em contrapartida, aos colegas acadêmicos e estudantes, especialmente os liga-dos aos campos da Administração de Empresas, Psicologia Social e Organizacional, assim como Sociologia, apresento farto material para reflexão já que empreendi ampla revisão literária, buscando coligir o estado da arte dos citados campos, bem como os resultados de pesquisas internacionais e nacionais. Nesse sentido, os leitores desse grupo encontrarão a análise de uma ampla gama de estudos abrigados em diferentes tópicos e cobrindo variados enfoques, metodologias e realidades investigadas. Eventualmente, forneço sugestões para futuros estudos.
Full-text available
We conduct a resume field experiment in all U.S. states to study how state laws protecting older workers from age discrimination affect age discrimination in hiring for retail sales jobs. We relate the difference in callback rates between old and young applicants to state variation in age and disability discrimination laws. These laws could boost hiring of older applicants, although they could have the unintended consequence of deterring hiring if they increase termination costs. In our preferred estimates that are weighted to be representative of the workforce, we find evidence that there is less discrimination against older men and women in states where age discrimination law allows larger damages, and more limited evidence that there is lower discrimination against older women in states where disability discrimination law allows larger damages. Our clearest result is that these laws do not have the unintended consequence of lowering callbacks for older workers.
Full-text available
This article presents the findings of the first field experiment - a resume correspondence study - on age discrimination in the Russian labor market. Correspondence studies are nowadays viewed as the most objective way to test for hiring discrimination. Th is method consists of sending pairs of CVs for job offers, very similar in everything except the trait to be analyzed (age in our case). Data collection for the presented study was conducted in February-March 2018. Pairs of matched applications, one from a fictitious 29-year-old female applicant and one from a fictitious 48-year-old female applicant, were sent to 341 employers with job openings for accountants in Moscow, posted on one of the most popular job search websites. It turned out that the probability of receiving an invitation for an interview for an older candidate is 24-32%, whereas for a younger candidate it is 45-52%. Th us, the positive callback ratio is 1.8-2.5. The indicator of "net discrimination", calculated as the difference in the shares of positive callbacks in the number of vacancies for which at least one response has been received, is 37-49% for the younger and ol der candidates. Compared with the results of the existing studies, it turns out that the obtained discrimination level is quite high. Taking into account the population aging, a reduction in the workforce and an increase in the official retirement age, measures must be taken to combat age discrimination and allow older people to work on an equal basis with young people.
This paper is about the reemployment difficulties of “mainstream” workers dislocated in mid-career by structural economic change. Contrary to popular assumptions, empirical evidence suggests that most workers in this category do not encounter extraordinary reemployment difficulties. For those that do, the root problem is not dislocation per se but rather such well-known failures of the private labor market as geographic immobility, underinvestment in training, and inefficient labor exchange institutions. Public policy should be structured in terms of correcting these market failures, not in terms of dislocated workers as atarget group.
In this paper we attempt to determine the degree to which older persons employed full-time experience a decline in earnings not related to a decline in productivity. We label such a decline in earnings “discrimination.” Using panel data for a cohort of men aged 45 to 54 in 1966, we examined trends in earnings between 1966 and 1976 and the effects of labor force experience on earnings controlling for factors related to productivity, such as health, time on the job, and specific occupational training, as well as other factors known to influence earnings levels. Comparing earnings attainment models estimated at both time periods, the effect of experience ori earnings was negligible in 1966 but substantial and negative in 1976. By far the largest component in the real earnings decline experienced by these men between 1966 and 1976 was associated with the increase in age-related experience.