Responsible researcher: Bruno Benevit
Original Title: Is It Harder for Older Workers to Find Jobs? New and Improved Evidence from a Field Experiment
Authors: David Neumark, Ian Burn and Patrick Button
Intervention Location: United States
Sample Size: 40,000 job applications
Sector: Labor Economy
Variable of Main Interest: Return of job applications
Type of Intervention: Age group
Methodology: Probit
Summary
The selection of workers may involve various forms of discrimination, notably related to issues of race or gender. However, in this context, economic literature has also given attention to discrimination due to age (ageism) in the hiring process. To examine this form of discrimination, this study applied a resume matching experiment in order to mitigate possible biases observed in previous studies. The results showed robust evidence of age discrimination in the hiring of older women, with great emphasis on the age group close to retirement age. Although signs of discrimination against men were also observed, the results were considerably less significant.
Older workers' longer unemployment spells have long been considered to potentially reflect hiring discrimination against older workers (NEUMARK; BURN; BUTTON, 2019). Given today's higher age populations, understanding and identifying this problem is fundamental. Several studies that analyzed the topic adopted the audit or correspondence (AC) methodology to verify the existence of this phenomenon.
Usually, studies that applied CA methods to evaluate this topic carry out a CV submission experiment, approximating the characteristics of the candidates, except age. However, this methodology disregards incentives faced by the group of older workers to reveal low experience on their CVs, information that may indicate a negative signal for this type of candidate. This situation indicates different behaviors between the group of younger and older workers, implying possible sources of bias when adopting such an approach. Therefore, to investigate the dynamics involved in this form of discrimination, it is necessary to consider specificities inherent to this problem.
Population aging is a phenomenon that has significant impacts on several aspects of society, including the labor market and social security systems. In many countries, such as the United States and many other places, the increasing proportion of elderly people in the population is creating fiscal challenges for Social Security systems. This is partly due to the lower employability of older workers, which results in a reduced tax base and, consequently, less revenue for social security programs. Additionally, this type of discrimination can be especially harmful when seniors want to transition to part-time jobs or return to work after a period of retirement.
In the context of the United States, the anti-discrimination law Age Discrimination in Employment Act (ADEA) establishes sanctions for discriminatory conduct against workers over 40 years of age. Simultaneously, there are economic costs associated with law enforcement compliance. According to the authors, it has been estimated that the costs of potential and actual contracting cases under the ADEA are about $3.29 billion per year, which equates to about $5,300 per ADEA-covered company, or about U.S. $35 per covered employee (NEUMARK; BURN; BUTTON, 2019).
To conduct the CA experiment, CVs were created covering several observable characteristics commonly associated with the probability of returning to job interviews. To establish parameters similar to those observed in reality, more than 25 thousand CVs were considered from a job search website. 12 cities were considered, selected by categories according to the prevalence of inhabitants aged 62 or over and the sanctions resulting from discriminatory practices in the hiring process. In total, the experiment carried out applications for more than 40 thousand job vacancies.
The CVs were created considering vacancies in occupations with a significant presence of young and senior workers with low experience according to the Current Population Survey (CPS). Additionally, the occupations with the highest prevalence were defined according to what was observed according to sex. The names and surnames displayed on the CVs were defined in such a way as to indicate that the competitors were Caucasian, aiming to isolate the analysis to the age aspect. Aiming to simulate a feasible scenario, the CVs presented three subjectively defined stories according to each city-occupation pair. Each summary was generated and submitted together with two other resumes differentiated by age (randomly), comparable experience or career transition moment described in the resume stories (beginning, middle and end of the career).
The study applied several models with the aim of verifying whether the age group of workers influences the probability of return. To this end, different specifications were considered based on the characteristics available in the CVs created and submitted in the AC experiment conducted, presenting estimates for all CVs and for each position-sex group.
Among the characteristics included in the vector of covariates of the models, the binary variables identifying the age range of the workers represented in the CVs were considered: (i) young workers (29-31 years old), (ii) middle-aged workers (49-51 years), and (iii) senior workers (64-66 years). The rest of the summary characteristics considered involve length of experience, gender, periods of career transition, completed higher education, fluency in Spanish, “employee of the month” highlight in the most recent job, participation in volunteer work, absence of typographical errors in the CV, two specific skills for the desired position, and the city of the desired position.
The main analysis considered a simple model, proposing to identify the existence of age discrimination directly by observing the return rates of CVs. Together, Fischer's independence test was applied in relation to the age of the candidates and the return rates. The purpose of the test was to verify the existence of a difference between the return rates between each of the three categories, comparing the three together (young vs middle-aged vs seniors) and the three possible pairs of categories (young vs middle-aged, young vs seniors, middle aged vs seniors).
Furthermore, Probit models with various specifications were estimated. Initially, a model was estimated with the aim of verifying the effects on the coefficients associated with each of the three age groups. Subsequently, a model was estimated comparing the results between older and younger candidates, depending on whether the CVs of older candidates show the same low experience as the CVs of younger candidates or, instead, experience commensurate with their age. . For this purpose, the presence of a difference between CVs with low experience and those with high experience/without career transition was verified.
The results of the main analysis showed strong evidence of discrimination between the young group and the older group (middle-aged and seniors). Overall, compared to young people's resumes, the return rates for middle-aged and senior resumes were 18% and 35% lower, respectively.
Specifically analyzing the return rates for each gender position, heterogeneity was observed in the degree of discrimination between the intended positions. The biggest difference was observed in applications from women for administrative positions, where 29% and 47% lower return rates were received for middle-aged and senior applications, respectively. Regarding male applications for security and janitorial positions, evidence of discrimination between young people and seniors of approximately 10% and 20%, respectively, was observed. The results regarding applications for sales positions, the only position considered for men and women, indicated discrimination and differences in terms of magnitude between the sexes. Compared to applications from young people for sales positions, 10% lower return rates were observed for middle-aged women and 35% lower for senior women, while for men differences were only identified in return rates for seniors (30% smaller compared to young people).
Overall, for all vacancies considered, the Probit model results revealed that return rates for older applicants were consistently lower than for younger applicants. Estimates of return rates for middle-aged applications were between 2.7 and 3.5 percentage points (pp) lower, while the coefficient associated with senior CVs indicated a reduction in the same rate of 5.4 to 6 .2pp. The estimates revealed that the results were not affected by the two types of differences in the professional trajectories of older candidates, nor by the experience measure considered, with the exception of applications for janitorial positions – in this case, being significant only in cases where the experience of seniors is the same as young people. The results considering a heteroscedastic Probit model were similar to those observed in the previous model, but maintaining the robustness of the results only for the coefficients associated with women.
This article verified the existence of age discrimination in the United States job market, observing the return of resumes from more than 40,000 job applications. From a matching experiment, CVs defined in three age groups (young, middle-aged and senior) were created to observe this type of dynamics, where the age group of the oldest group was close to retirement ages (64 -66 years).
The results found in this study indicate the existence of age discrimination, so that CVs from younger candidates were more likely to receive feedback compared to middle-aged and, to a greater extent, senior candidates. Furthermore, it was observed that the positions analyzed for women presented more prevalent and higher discriminatory impacts when compared to positions for men. The evidence from this study highlights the importance of anti-discrimination policies and contributes to the understanding of how ageism occurs in the labor market, mainly impacting workers close to retirement age.
References
NEUMARK, D.; BURN, I.; BUTTON, P. Is It Harder for Older Workers to Find Jobs? New and Improved Evidence from a Field Experiment. Journal of Political Economy , vol. 127, no. 2, p. 922–970, apr. 2019.
Check the Institution's registration in the e-MEC System here
© 2025 All Rights Reserved
