Who Gets Hired? The Importance of Competition among Applicants
GAICIL Insights
A job offer is a verdict delivered by comparison.1 Who Gets Hired? The Importance of Competition among Applicants, by Edward P. Lazear, Kathryn L. Shaw and Christopher T. Stanton, makes the point rigorous: being hired depends on the skill of other applicants as well as on one's own. When another able applicant shows up, a well-suited worker may be pushed into unemployment or into an inferior job.
The model is built on job slots. Firms post positions, and even qualified workers may miss out, or land in the wrong role, because a superior applicant competes for the same slot. The model defines overqualification and underqualification and predicts their prevalence and the wages of mismatched workers. It implies that unemployment concentrates among the least skilled while vacancies concentrate among high-skilled jobs.
Four data sets test the predictions. The probability of being hired depends on the skills of the competition, verified with oDesk data from an online labor market. Bumping occurs: a worker takes a job he is not well suited for because his skills exceed every other applicant's except the one who takes his preferred job. PSID data show overqualification and underqualification occur, with wages in mismatched jobs following the model's predictions. Both overqualified and underqualified workers earn less than they would in appropriate positions.
Bumping is the image to remember: a cascade down a ladder. One strong applicant arrives at the top rung, and everyone below slides down a step, until the person at the bottom falls off.
AI can start that cascade. AI tools raise the effective skills of some workers, making them stronger competitors for particular jobs. Workers who use AI well may outcompete others, bumping them down the ladder. The model predicts such shifts concentrate unemployment among the least skilled and reshape mismatch across the labor market. The displaced worker may never have been replaced by a machine at all; she was outcompeted by a colleague who uses one.
AI also rewires matching itself. Online labor markets of the kind in the paper's oDesk analysis increasingly use algorithms to pair applicants and jobs. Better matching could reduce mismatch by helping workers find the positions that suit them and helping firms spot the right candidates. The model supplies yardsticks: falling overqualification and underqualification, and shrinking wage penalties, would signal more efficient matching. Those are measurable promises AI matching platforms can be held to.
The concentration of unemployment at the bottom carries a clear policy message. If AI lifts many workers' skills, competition for jobs intensifies, and the least skilled face the greatest risk of being bumped into unemployment. Raising skills at the bottom, including training in AI tools, helps them compete. Broad access to skill-enhancing technology could narrow differences among applicants and reduce the concentration of unemployment. Equal access to the tool levels the competition for the slot.
The vacancy side matters too. If AI creates high-skilled roles faster than workers can acquire the skills, vacancies will linger there. Education and training systems that respond quickly to shifting skill demands can fill them and let the economy capture AI's productivity gains more fully. An unfilled AI-era job is lost output.
Job slots themselves are not fixed. AI may change the structure of jobs as well as the skills of workers. New slots defined around AI-assisted tasks will open new competitions among applicants, and understanding how those play out requires exactly the kind of model and evidence the authors provide.
At bottom, labor market outcomes are relative. A worker's prospects depend on everyone else in the pool. As AI reshapes the distribution of skills, it reshapes the competition for jobs. Helping every worker become a stronger applicant, through skills and access to technology, offers the best route to a labor market where the gains from AI are widely shared.
The provocative implication for policy: the most effective protection against AI-driven unemployment may be ensuring that the workers at the bottom of the ladder get the same tools as those at the top.
Introduction
Being hired into a job depends not only on one’s own skill but also on that of other applicants. When another able applicant applies, a well-suited worker may be forced into unemployment or into accepting an inferior job. A model of this process defines over- and underqualification and provides predictions on its prevalence and on the wages of mismatched workers. It also implies that unemployment is concentrated among the least skilled workers, while vacancies are concentrated among high-skilled jobs. Four data sets are used to confirm the implications and establish that the hiring probability is low when competing applicants are able.
Takeaway
“A worker’s skills alone do not determine the job into which he or she is hired or, indeed, whether the worker is hired at all. The existence of job slots that firms post means that even qualified workers may not be hired or may not be assigned to the job for which they are best suited because there is a superior applicant for the posted position. Although this idea is intuitive, it has not been modeled in a way that allows its implications to be explored in individual-based data. The model and analysis presented herein not only provides many specific predictions on what should be observed in hiring and job assignment but also tests and validates those predictions using four different data sets. First, the probability of being hired depends not only on an applicant’s skills but also on the skills of the competition. This is verified using oDesk data. Second, bumping occurs, which happen when a worker takes a job for which he is not well suited but receives the offer because his skills are superior to those of other applicants except the one who gets the job that he prefers. The model provides clear definitions of “overqualification” and “underqualification” that have specific empirical meaning. Using these definitions, the PSID data provide evidence that over-and underqualification occurs and that the wages that are received in those jobs are consistent with the predictions of the model.1
Specifically, overqualified workers receive lower pay in the job for which they are overqualified than they would in their appropriate positions. Underqualified workers receive lower pay in the job for which they are underqualified than they would in their appropriate positions, but the results for the underqualified are not statistically significant.”
Notes
- Edward P. Lazear, Kathryn L. Shaw & Christopher T. Stanton, Who Gets Hired? The Importance of Competition among Applicants, 36 J. Lab. Econ. S133 (2018), https://doi.org/10.1086/694908. Takeaway source: Edward P. Lazear, Kathryn L. Shaw, Christopher T. Stanton, Who Gets Hired? The Importance of Competition among Applicants, Conclusion. ↩
Cite as: Edward P. Lazear, Kathryn L. Shaw & Christopher T. Stanton, Who Gets Hired? The Importance of Competition among Applicants, 36 J. Lab. Econ. S133 (2018), https://doi.org/10.1086/694908
Header image: 003 Car dashboard manufacturing process - Opel car factory in Gliwice, Poland.jpg by Marek Ślusarczyk (Tupungato) Photo portfolio, CC BY 3.0, via Wikimedia Commons.
