# arXiv cs.AI: Human Capital in Context: What Employment Survival Reveals About Worker-Firm Match [Canonical reading](https://escapemeridian.com/readings/arxiv-cs-ai-human-capital-in-context-what-employment-survival-edf5f0c5) Jung Ho Choi and Mario Milone argue that there is no universally best worker: the value of a worker depends on the firm. They measure worker-firm match using a two-tower neural network, which represents workers and firms in a shared space; match quality is the cosine similarity of their representations. The model learns match quality from employment separation, treating a job match as an experience good. Using more than a billion person-periods, the authors report that a 0.1 increase in match quality lowers the odds of separation by 48%, and that about a quarter of match-quality variation is specific to a worker-firm pair. They report that better-matched workers earn more, especially early in a relationship. Firms with higher-scoring workforces grow faster, become more profitable, and receive more and more valuable patents over the next five years than industry peers. Because the measure can score pairs that never formed, the authors say it predicts where workers move. They also report that acquirers record a larger share of the price as goodwill when the target’s workers fit the acquirer better than they fit the target. Escape Meridian interpretation: the abstract does not explain how the analysis separates match quality from other reasons jobs persist or firms perform well. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 11, 2026 (2026-11-11). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-13T13:50:15.386Z. First observed: 2026-11-13T07:43:43.044Z. Retrieved: 2026-11-13T13:48:09.219Z. - [Human Capital in Context: What Employment Survival Reveals About Worker-Firm Match](https://arxiv.org/abs/2611.16058) ## Limits The supplied evidence is the arXiv page and abstract, not the paper's full methods or results; it does not describe how the analysis distinguishes match quality from other factors associated with employment survival. The reported relationships do not, from the abstract alone, establish that match quality causes earnings, firm growth, profitability, patenting, worker moves, or acquisition goodwill. ## Open question How does the measure distinguish match quality from other reasons employment relationships persist? ## Source-stated Choi and Milone define match quality as a property of a worker-firm pair, rather than a worker's general quality. > There is no best worker in general, only a best worker for a particular firm. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors estimate match quality by representing workers and firms in a shared space and calculating the cosine similarity of their vectors. > A two-tower neural network embeds workers, from their career histories and current positions, and firms, from their characteristics and business descriptions, into a common space, and match quality is the cosine similarity of the two vectors. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated In the authors' estimate, a 0.1 increase in match quality lowers the odds of separation by 48%. > a 0.1 increase in match quality lowers the odds of separation by 48%. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors report that about a quarter of match-quality variance is specific to the worker-firm pair. > About a quarter of the variance of match quality is specific to the worker-firm pair - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors report that better-matched workers earn more, particularly early in a relationship, and that firms with higher-scoring workforces outperform industry peers on several reported measures over the next five years. > Workers earn more when better matched, especially early in a relationship, and firms whose workforces score higher grow faster, become more profitable, and are issued more, and more valuable, patents over the next five years than their industry peers. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Escape Meridian interpretation The abstract reports findings based on employment survival but does not explain how the analysis separates match quality from other factors associated with employment persistence. > its only estimated objective, a separation hazard, from the view of the job match as an experience good. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source and contributor context Choi, Jung Ho, Milone, Mario · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Neural network A system of connected simple calculations, loosely inspired by the brain, that learns patterns from data. - [Definition reference](https://developers.google.com/machine-learning/glossary#neural-network) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 33e5cdc9-87bb-4631-80d1-bf6f7a9cd510. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T13:50:15.386Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: e0f7c98e1284f3d82d88b8f6f672eb2bfb3a24757f253714c5948e042673b801. Source text hash: 4dd3a15b267a7e646b4e755e05f04923b533a79ea6432ee5cc63885382e7746b. ## Revision and source-check history ### Version 1 — 2026-11-13T13:50:15.386Z Jung Ho Choi and Mario Milone argue that there is no universally best worker: the value of a worker depends on the firm. They measure worker-firm match using a two-tower neural network, which represents workers and firms in a shared space; match quality is the cosine similarity of their representations. The model learns match quality from employment separation, treating a job match as an experience good. Using more than a billion person-periods, the authors report that a 0.1 increase in match quality lowers the odds of separation by 48%, and that about a quarter of match-quality variation is specific to a worker-firm pair. They report that better-matched workers earn more, especially early in a relationship. Firms with higher-scoring workforces grow faster, become more profitable, and receive more and more valuable patents over the next five years than industry peers. Because the measure can score pairs that never formed, the authors say it predicts where workers move. They also report that acquirers record a larger share of the price as goodwill when the target’s workers fit the acquirer better than they fit the target. Escape Meridian interpretation: the abstract does not explain how the analysis separates match quality from other reasons jobs persist or firms perform well. ## Original source arXiv cs.AI · Article Original source date: Nov 11, 2026 (2026-11-11). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-13T13:50:15.386Z. First observed: 2026-11-13T07:43:43.044Z. Retrieved: 2026-11-13T13:48:09.219Z. - [Human Capital in Context: What Employment Survival Reveals About Worker-Firm Match](https://arxiv.org/abs/2611.16058) ## Limits The supplied evidence is the arXiv page and abstract, not the paper's full methods or results; it does not describe how the analysis distinguishes match quality from other factors associated with employment survival. The reported relationships do not, from the abstract alone, establish that match quality causes earnings, firm growth, profitability, patenting, worker moves, or acquisition goodwill. ## Open question How does the measure distinguish match quality from other reasons employment relationships persist? ## Source-stated Choi and Milone define match quality as a property of a worker-firm pair, rather than a worker's general quality. > There is no best worker in general, only a best worker for a particular firm. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors estimate match quality by representing workers and firms in a shared space and calculating the cosine similarity of their vectors. > A two-tower neural network embeds workers, from their career histories and current positions, and firms, from their characteristics and business descriptions, into a common space, and match quality is the cosine similarity of the two vectors. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated In the authors' estimate, a 0.1 increase in match quality lowers the odds of separation by 48%. > a 0.1 increase in match quality lowers the odds of separation by 48%. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors report that about a quarter of match-quality variance is specific to the worker-firm pair. > About a quarter of the variance of match quality is specific to the worker-firm pair - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source-stated The authors report that better-matched workers earn more, particularly early in a relationship, and that firms with higher-scoring workforces outperform industry peers on several reported measures over the next five years. > Workers earn more when better matched, especially early in a relationship, and firms whose workforces score higher grow faster, become more profitable, and are issued more, and more valuable, patents over the next five years than their industry peers. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Escape Meridian interpretation The abstract reports findings based on employment survival but does not explain how the analysis separates match quality from other factors associated with employment persistence. > its only estimated objective, a separation hazard, from the view of the job match as an experience good. - [Supporting source](https://arxiv.org/abs/2611.16058) ## Source and contributor context Choi, Jung Ho, Milone, Mario · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Neural network A system of connected simple calculations, loosely inspired by the brain, that learns patterns from data. - [Definition reference](https://developers.google.com/machine-learning/glossary#neural-network) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 33e5cdc9-87bb-4631-80d1-bf6f7a9cd510. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T13:50:15.386Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: e0f7c98e1284f3d82d88b8f6f672eb2bfb3a24757f253714c5948e042673b801. Source text hash: 4dd3a15b267a7e646b4e755e05f04923b533a79ea6432ee5cc63885382e7746b. ### uncertain — 2026-11-13T13:50:15.386Z — anthropic/claude-opus-5.5 Publication: 33e5cdc9-87bb-4631-80d1-bf6f7a9cd510 attribution — pass: The reading correctly attributes the argument and findings to authors Jung Ho Choi and Mario Milone in an arXiv paper. Results are framed as the authors' reports, not as established facts. > Authors:Jung Ho Choi, Mario Milone summary — uncertain: Mostly faithful, but two gaps remain. First, central terms go unexplained for a general reader: 'experience good,' 'goodwill,' and 'person-periods.' The 'experience good' rationale is never explained. Smallest repair: add tooltips or short glosses, and suggest term-library entries. Second, the 'about a quarter' finding drops the paper's key point. The source says worker-, firm- and occupation-level measures set this pair-specific share to zero by construction, which is why the pair matters. Restoring that clause would recover the argument's force. > About a quarter of the variance of match quality is specific to the worker-firm pair, which worker-, firm-, and occupation-level measures set to zero by construction. date — pass: The originalPublished date of 2026-11-11 matches the arXiv v1 submission date. It is not the later feed date. > \[Submitted on 11 Nov 2026\] support — pass: The claims match the abstract: the two-tower model, the cosine similarity measure, the 48% odds reduction, the pair-specific variance, earnings and firm outcomes, mover prediction, and goodwill. The Escape Meridian interpretation is labeled and conditional. It rests on the abstract's silence about how match quality is separated from other causes. The limitations properly avoid causal claims. The question follows from the evidence without leading. > its only estimated objective, a separation hazard, from the view of the job match as an experience good. media — pass: The selected media is the verified direct link to the original arXiv abstract page. It is the most relevant source access. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/edf5f0c5b56ca51c21c2d1f5cbc74e5b52e217ad37207fc2dc1585e72b909bef/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/arxiv-cs-ai-human-capital-in-context-what-employment-survival-edf5f0c5.json)