Great Teamwork
The Dynamics of Exploiting Overconfident Workers
Matthias Fahn & Nicolas Klein
Journal of Labor Economics, forthcoming
Abstract:
We develop a model of a long-term employment relationship in which a firm may have an incentive to reassign a worker who has proved himself to be good at his current job — even if he is not well-suited for the new position. This occurs because, once a worker’s talent becomes evident, the firm can no longer exploit his overconfidence, which reduces the profitability of his current employment. This insight provides a new microfoundation for the (intentional) mismatch between employees and their tasks that is commonly referred to as the “Peter Principle.”
Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data
Seyed Hosseini & Guy Lichtinger
Harvard Working Paper, October 2025
Abstract:
We study whether generative AI (GenAI) constitutes seniority-biased technological change, disproportionately affecting junior relative to senior workers. Using U.S. résumé data covering 62 million workers across 285,000 firms (2015-2025), we track firm-level employment by seniority. GenAI adoption is identified through text analysis that detects “GenAI integrator” job postings, signaling active GenAI implementation by firms. Adoption is associated with a sharp decline in junior employment among adopting firms relative to non-adopters, while senior employment remains largely unchanged. The junior decline is concentrated in occupations most exposed to GenAI and is driven by slower hiring rather than increased separations or promotions.
When Autonomy Backfires: Adverse Selection in Startup Recruitment
Saerom (Ronnie) Lee & Reuben Hurst
University of Pennsylvania Working Paper, August 2025
Abstract:
While prior entrepreneurship research highlights strong preferences for autonomy among founders and early joiners of startups, little is known about how explicitly emphasizing autonomy in startup recruitment affects the composition of the applicant pool. We propose that such emphasis may inadvertently reduce applicant quality by disproportionately attracting lower-ability candidates while deterring higher-ability job seekers. This adverse selection may arise because individuals with varying cognitive abilities may interpret autonomy-related signals in systematically different ways. Specifically, lower-ability candidates may perceive autonomy at face value as desirable freedom from managerial oversight, whereas higher-ability individuals may infer potential organizational dysfunction, including ill-defined roles, insufficient managerial oversight, and unclear performance evaluation. Using large-scale job posting data, we first document that startups increasingly highlight autonomy in recruitment, particularly when targeting higher-ability talent. We then provide causal evidence for the proposed adverse selection through a field experiment conducted in partnership with a U.S.-based startup. Our findings reveal that, although autonomy is widely considered a desirable attribute that distinguishes new ventures from established firms, its explicit emphasis in startup recruitment may backfire -- paradoxically deterring the high-ability candidates most critical to entrepreneurial success.
AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights
Jiannan Xu, Gujie Li & Jane Yi Jiang
University of Maryland Working Paper, September 2025
Abstract:
As generative artificial intelligence (AI) tools become widely adopted, large language models (LLMs) are increasingly involved on both sides of decision-making processes, ranging from hiring to content moderation. This dual adoption raises a critical question: do LLMs systematically favor content that resembles their own outputs? Prior research in computer science has identified self-preference bias -- the tendency of LLMs to favor their own generated content -- but its real-world implications have not been empirically evaluated. We focus on the hiring context, where job applicants often rely on LLMs to refine resumes, while employers deploy them to screen those same resumes. Using a large-scale controlled resume correspondence experiment, we find that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, even when content quality is controlled. The bias against human-written resumes is particularly substantial, with self-preference bias ranging from 68% to 88% across major commercial and open-source models. To assess labor market impact, we simulate realistic hiring pipelines across 24 occupations. These simulations show that candidates using the same LLM as the evaluator are 23% to 60% more likely to be shortlisted than equally qualified applicants submitting human-written resumes, with the largest disadvantages observed in business-related fields such as sales and accounting. We further demonstrate that this bias can be reduced by more than 50% through simple interventions targeting LLMs' self-recognition capabilities. These findings highlight an emerging but previously overlooked risk in AI-assisted decision making and call for expanded frameworks of AI fairness that address not only demographic-based disparities, but also biases in AI-AI interactions.
Are less hierarchical firms organized around stronger cultures? Evidence from big data
Arianna Marchetti & Phanish Puranam
Strategic Management Journal, forthcoming
Abstract:
Are less hierarchical firms organized around stronger cultures instead? We analyze 1.5 million employee reviews on Glassdoor.com from 23,000 US-based firms, alongside data on managerial hierarchy estimated from 42 million professional social media profiles. Our findings confirm a negative association between managerial hierarchy and organizational culture strength. We explore two potential explanations for this association: Functional equivalence between the two, and culture fragmentation caused by managerial hierarchy. Multiple correlational tests show support for functional equivalence as a plausible explanation for the observed negative correlation. Our findings enhance our understanding of the complex relationships between organizational structure and culture utilizing Big Data methods.
Dissecting Corporate Culture Using Generative AI
Kai Li et al.
Review of Financial Studies, forthcoming
Abstract:
We conduct the first large-scale study of how different stakeholder groups assess corporate culture and quantify the economic implications of those differences. We employ generative AI to analyze analyst reports, call transcripts, and employee reviews, and organize the extracted information into a knowledge graph that links a culture type to its perceived causes and effects. We demonstrate that the divergence in different stakeholder groups’ assessment of culture aligns with their distinct roles and economic incentives. Moreover, we show that analysts’ culture analyses are incorporated into stock recommendations and target prices, and investors react to divergence in stakeholders’ assessment of culture.
When and Why “Consoling” Marginal Underperformers With a Small Versus Zero Reward Hurts Fairness (Without Consolation)
Minzhe Xu & Bowen Ruan
Journal of Behavioral Decision Making, October 2025
Abstract:
When resource allocation decisions involve marginal underperformers (MUs) -- individuals or parties who underperform only “by inches” relative to a threshold -- allocators may adopt the consolatory approach, compensating MUs with a small portion of the total resource. Seven studies (N = 2585) revealed that the consolatory approach, albeit often well intended, may backfire. Specifically, when compared with the non-consolatory, binary approach (allocating all the resource to outperformers and nothing to MUs), the consolatory approach can be perceived as less fair, even by MUs themselves who economically benefit from it. As the consolatory approach is objectively more equitable than the binary approach, this effect contradicts the prediction of proportional equity, thereby demonstrating its discontinuity at zero. The underlying mechanism is grounded in people's fundamental perception of zero as unique relative to other numbers, which leads them to adopt different criteria to evaluate fairness depending on whether the allocation outcomes involve zero. This work suggests that the common practice of offering MUs a small “consolation prize” may backfire, harming fairness without mitigating MUs' negative feelings of losing.
“Zoom fatigue” revisited: Are video meetings still exhausting post-COVID-19?
Hadar Nesher Shoshan & Wilken Wehrt
Journal of Occupational Health Psychology, forthcoming
Abstract:
During COVID-19, participating in video meetings was associated with exhaustion, a phenomenon termed “Zoom fatigue.” Different explanations were suggested for this phenomenon. Video meetings may be exhausting because they (a) gained a symbolic meaning during the pandemic, (b) are cognitively demanding, or (c) are understimulating, leading to “passive fatigue.” Characterizing video meetings as “fatiguing” might have critical implications for hybrid work policy. Therefore, we must ask whether video meetings are still exhausting post-COVID-19. In 2024, we constructively replicated an experience sampling study about “Zoom fatigue” during the pandemic. Although the meaning of video meetings might have changed, considering the current state of the literature, we hypothesized that ”Zoom fatigue” still exists. We conducted a 10-day experience sampling study with four daily measurement points. Results from a three-level analysis (N = 125participants, n = 590days, n = 945meetings) showed that video meetings are unrelated to exhaustion or passive fatigue. Active participation in the meeting and multitasking were not significant moderators. Video meetings shorter than 44 min were less exhausting than other meetings, suggesting potential advantages of video meetings. More boring video meetings were slightly more exhausting. Our results highlight the importance of constructively replicating research findings in different historical settings. They support the view that video meetings gained a symbolic meaning during the pandemic, which may have changed afterward. We also suggest that people might have gotten used to video meetings, leading to “no-Zoom fatigue.” Our results open avenues for future studies and have important practical implications.
Temporary employment and the protection of investments in human capital: Examining the Major League Baseball player market
Richard Paulsen
Economic Inquiry, October 2025, Pages 1279-1290
Abstract:
When employees are employed in a temporary capacity, employers should be less willing to invest in their human capital relative to permanent employees. This study uses the context of injury management by Major League Baseball teams to test for differential investment in the protection of player human capital. Injury management is inherently uncertain as medical professionals can give differing opinions, so teams may be able to influence recovery times. Using a panel dataset and estimating player fixed-effects regressions, players are found to miss significantly fewer games to injury when employed on a temporary basis.
The Lower Boundary of Workplace Mistreatment: Do Small Slights Matter?
Michal Hodor, Liat Eldor & Peter Cappelli
NBER Working Paper, October 2025
Abstract:
Recent research in psychology, management, and more recently in economics, highlights the role of individual managers and their behavior in shaping employee performance. While emerging literature on harmful managerial behavior has focused primarily on severe forms of workplace mistreatment, especially various types of harassment, much less is known about its boundary conditions: How minor can a manager’s bad behavior be and still negatively affect employee performance? We study what appears to be a very minor workplace mistreatment -- failing to deliver an expected birthday gift and greeting card on time -- and examine its effect on subsequent employee performance. Using a dynamic difference-in-differences approach with detailed data from a national retail chain, we find that this small slight leads to over a 50% increase in employee absenteeism and a reduction of more than two working hours per month. Our analysis suggests that emotional responses to perceived workplace mistreatment drive the results. These findings indicate that even modest slights can meaningfully harm employee performance.