Time Travel on Professional Profiles, with Nicholas Bloom, Gideon Moore and Caelan Wilkie-Rogers, NBER Working Paper No. 35546, August 2026.
Economists increasingly use professional profile data to reconstruct employment histories and measure skill supply. We show that these records are not fixed historical snapshots, but mutable accounts that workers revise over time. Using monthly vintages of Revelio Labs data from 2020–2026, we document that 19.7 percent of established US LinkedIn users retroactively edit the title or description of a job they have already left. These “time-travel” edits are closely tied to labor market transitions: around such edits, workers are much more likely to change employers as compared to later-editing users. This mutability can bias historical measures of skills, but it also reveals workers’ beliefs about which skills are in demand. Retroactive edits show sharp post-2022 increases in AI-related language and recent reductions in work-from-home and DEI language. Finally, LLM-associated writing markers surge after ChatGPT, especially among less-educated groups and MBAs from lower-ranked programs, revealing heterogeneous AI-assisted profile editing.
A New Look at AI’s Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment, with Ara Kharazian and Ryan Stevens, SSRN Working Paper, June 2026.
We study how employment changes when firms adopt generative AI using observed AI spending from Ramp card and bill pay data linked to Revelio Labs workforce records for 21,559 firms in the United States. We find that companies that adopt AI tend to grow faster following adoption, but the relationship is driven almost entirely by high-intensity adopters. Firms making the largest AI investments grow employment by roughly 10% following adoption, while low-intensity adopters see no statistically significant change. Entry-level headcount rises 12% for high-intensity adopters. Gains emerge gradually and are broad across roles, including engineering, sales, administration, and customer service. They are also uneven: adopters are already larger, more technical, faster-growing firms, and sector-level gains are concentrated in Information. The results counter predictions that AI adoption will lead to broad job loss.
The Heterogeneous Earnings Impact of Job Loss Across Workers, Establishments, and Markets, with Susan Athey, Oskar Nordström Skans, Johan Vikström and Yaroslav Yakymovych, NBER Working Paper No. 34946, March 2026.
Using rich Swedish administrative data, we apply causal machine learning methods to study how earnings losses after job displacement vary with observable characteristics that may be relevant for targeting policy interventions for workers. Heterogeneity in effects is as large within as across worker groups defined by age and schooling, and as large within as across establishments. A substantial portion of cross-establishment heterogeneity can be explained by industry and local labor market characteristics, suggesting a role for place- and industry-based targeting. The largest losses are concentrated among already vulnerable workers, indicating that well-designed targeting policies can improve both efficiency and equity.
Earlier version: arXiv:2307.06684, July 2023
AI-Exposed Jobs Deteriorated Before ChatGPT, with Morgan R. Frank, Alireza Javadian Sabet, Sarah H. Bana and Renzhe Yu, arXiv:2601.02554, January 2026.
Public debate links worsening job prospects for AI-exposed occupations to the release of ChatGPT in late 2022. Using monthly U.S. unemployment insurance records, we measure occupation- and location-specific unemployment risk and find that risk rose in AI-exposed occupations beginning in early 2022, months before ChatGPT. Analyzing millions of LinkedIn profiles, we show that graduate cohorts from 2021 onward entered AI-exposed jobs at lower rates than earlier cohorts, with gaps opening before late 2022. Finally, from millions of university syllabi, we find that graduates taking more AI-exposed curricula had higher first-job pay and shorter job searches after ChatGPT. Together, these results point to forces pre-dating generative AI and to the ongoing value of LLM-relevant education.
Adjusters and Casualties: The Anatomy of Labor Market Displacement, with Eric A. Hanushek, Simon Janssen and Jacob D. Light, NBER Working Paper No. 33667, May 2025.
We analyze the full distribution of displaced workers’ earnings losses using a new method that combines matching and synthetic control group approaches at the individual level. We find that the distribution of earnings losses is highly skewed. Average losses, as estimated by conventional event studies, are driven by a small number of workers who suffer catastrophic losses, while most recover quickly. Observable worker characteristics explain only a small fraction of the variance in earnings losses. Instead, we find substantial heterogeneity in earnings losses even among workers displaced by the same firm who have identical observed characteristics such as education, age, and gender. Workers with minimal earnings losses adjust quickly by switching industries, occupations, and especially regions, while comparable workers with catastrophic losses adjust slowly, even though they are forced to make comparable numbers of switches in the long run.
VoxEU column: Revisiting the consequences of job displacement, August 2025
Multidimensional Skills on LinkedIn Profiles: Measuring Human Capital and the Gender Skill Gap, with David Dorn, Florian Schoner, Moritz Seebacher and Ludger Woessmann, IZA Discussion Paper No. 17896, May 2025.
Earlier version: arXiv:2409.18638, September 2024
Shocking Choice: Trade Shocks, Local Labour Markets and Vocational Occupation Choices
Whether individuals choose occupations that teach them general or specific skills can have important implications on how protected they are to changing conditions on the labor market. This paper looks at the impact of growing up in German regions that are exposed to structural change on individual vocational occupation choices. Exogenous variation in local labor market structural change stems from shift-share type import shocks from China and Eastern Europe, which are instrumented by trade with other high income countries. Using longitudinal administrative social security data, results show that individuals surprisingly chose more skill-specific occupations as a result of higher local import competition. Lifetime earnings are adversely affected by being exposed to import competition during adolescence, which can be explained by occupational choice. The results are not driven by labor demand, nor endogenous sorting into vocational education.
Download ifo Working Paper 281, December 2018
Entry Barriers and the Labor Market Outcomes of Incumbent Workers: Evidence from a Deregulation Reform in the German Crafts Sector, with Philipp Lergetporer and Jens Ruhose
We study the labor market outcomes of a deregulation reform in Germany that removed licensing requirements to become self-employed in some occupations. Using longitudinal social security data, we implement a matched difference-in-differences design with entropy balancing to account for observable characteristics and unobserved individual heterogeneity. The reform tripled the number of businesses within ten years and led to slower earnings growth and higher unemployment for incumbent workers in deregulated occupations. However, the reform effect seems rather small, which we attribute to the relatively low competitiveness of new businesses. Supporting this view, the reform did not lead to overall employment growth.
Download IZA Working Paper, IZA DP No. 11857, September 2018
Beyond Surveys: How Alternative Data Can Measure the Health of the Labor Market, with Loujaina Abdelwahed and Ritesh Baldva, EconPol Forum 27(1), 58–65, 2026.
The private sector offers reliable labor market data that policymakers, investors, and the public can use to guide their decisions. Revelio Labs provides monthly labor statistics from billions of work records, tracking employment, openings, turnover, and pay. Alternative data is created beyond the goal of measurement, thereby avoiding survey bias and offering timely, cost-effective insights.
Can Internet Surveys Represent the Entire Population? A Practitioners’ Analysis, with Elisabeth Grewenig, Philipp Lergetporer, Katharina Werner and Ludger Woessmann, European Journal of Political Economy 78, 102382, 2023.
A general concern with the representativeness of internet surveys is that they exclude the “offline” population that does not use the internet. We run a large-scale opinion survey with (1) onliners in internet-survey mode, (2) offliners in face-to-face mode, and (3) internet users in face-to-face mode. We find marked response differences between onliners and offliners in different modes (1 vs. 2). Response differences between onliners and offliners in the same face-to-face mode (2 vs. 3) disappear when controlling for background characteristics, indicating mode effects rather than unobserved population differences. Differences in background characteristics of onliners in the two modes (1 vs. 3) indicate that mode effects partly reflect sampling differences. In our setting, re-weighting online-survey observations appears a pragmatic solution when aiming at representativeness for the entire population.
Older Version: CESifo Working Paper, No. 7222, September 2018
What Outperformers Do Differently to Tap Internal Talent, with Nithya Vaduganathan, Ben Zweig and Colleen McDonald, MIT Sloan Management Review 64(1), 2022.
Offering employees lateral moves can be an untapped gold mine for companies. By adopting leading practices for internal mobility, companies can better deploy existing worker capacity and benefit from more successful hires with greater institutional knowledge, higher engagement and retention, and improved gender equity.
Does the Education Level of Refugees Affect Natives’ Attitudes?, with Philipp Lergetporer and Marc Piopiunik, European Economic Review 134, 103710, 2021.
In recent years, Europe has experienced a large influx of refugees. While natives’ attitudes toward refugees are decisive for the political feasibility of asylum policies, little is known about how these attitudes are shaped by information about refugees’ characteristics. We conducted a survey experiment with a representative sample of more than 4,000 adults in Germany in which we randomly provide information about refugees’ education level. Information provision strongly increases respondents’ beliefs that refugees are well educated. The information also increases labor market competition concerns, decreases fiscal burden concerns, and positively affects general attitudes toward refugees. We perform several robustness analyses in additional experiments with more than 5,000 university students. In sum, we show that correcting misperceptions about refugees’ education level has profound effects on natives’ attitudes.
ifo Working Paper Version, No. 346, January 2021
Older Version: CESifo Working Paper, No. 6832, December 2017
Skills, Signals, and Employability: An Experimental Investigation, with Marc Piopiunik, Guido Schwerdt and Ludger Woessmann, European Economic Review 123, 103374, 2020.
As skills of labor-market entrants are usually not directly observed by employers, individuals acquire skill signals. To study which signals are valued by employers, we simultaneously and independently randomize a broad range of skill signals on pairs of resumes of fictitious applicants among which we ask a large representative sample of German human-resource managers to choose. We find that signals in all three studied domains – cognitive skills, social skills, and maturity – have a significant effect on being invited for a job interview. Consistent with the relevance, expectedness, and credibility of different signals, the specific signal that is effective in each domain differs between apprenticeship applicants and college graduates. While GPAs and social skills are significant for both genders, males are particularly rewarded for maturity and females for IT and language skills. Older HR managers value school grades less and other signals more, whereas HR managers in larger firms value college grades more.
Download CESifo Working Paper, No. 6858, January 2018
Health Care Coverage in OECD Countries in 2012, with Valérie Paris, Emily Hewlett, Ane Auraaen and Jan Alexa, OECD Health Working Papers No. 88, 2016.
This paper provides a detailed description of health coverage in OECD countries in 2012. It includes information on the organisation of health coverage (residence-based vs contributory systems), on the range of benefits covered by basic health coverage and on cost-sharing requirements. It also describes policies implemented to ensure universal health coverage –in most countries- and to limit user charges for vulnerable populations or people exposed to high health spending. The paper then describes the role played by voluntary health insurance as a secondary source of coverage. Combining qualitative information collected through a survey of OECD countries on benefits covered and cost-sharing requirements with spending data collected through the system of health accounts for 2012, this paper provides valuable information on health care coverage in OECD countries at a time universal health coverage is high on the policy agenda of many countries.
Download OECD Health Working Papers, No. 88, 2016