.svg)
Job posting data for labor market research
Job postings record labor demand at the moment a firm decides to hire, naming the occupation, the location, the skills required and, often, the offered pay, before any of it shows up in employment statistics. Because each posting is tied to a named employer, researchers can build firm level hiring panels that household surveys and establishment surveys cannot produce, and track how quickly a vacancy is filled or withdrawn. Skill text makes it possible to measure the adoption of a technology by the human capital firms buy rather than by capital expenditure. The tradeoff is that postings cover advertised vacancies rather than all hiring, so occupations and firms that recruit informally are underrepresented.
How labor economists use job posting data
Measuring technology adoption through hiring demand
Skill and title text in postings lets researchers measure which firms are adopting a technology and how fast, using the share of vacancies requiring named tools or competencies as a continuous exposure measure, then comparing outcomes across firms whose pre period occupational mix made them more or less exposed. Because the classification depends on keyword or model based tagging of free text, results move with the classifier: definitions of an AI role built from a fixed keyword list and from a language model applied to the same postings do not agree on which vacancies count. Postings also capture intended hiring rather than completed hires, so the measure leads actual headcount change by an unknown and probably variable lag.
Firm responses to regulatory and accounting change
Because postings identify the employer and the occupation, researchers can isolate hiring within a single corporate function, such as a tax or internal accounting department, and use a staggered or one time rule change as the shock in a difference in differences design with unaffected functions inside the same firm as the control. The variable is usually the count of postings in the treated occupation, sometimes paired with changes in the skills requested, which is what separates a firm adding headcount from a firm rewriting the job. Function level identification rests on occupation coding and on title text, and small departments generate thin counts, so the design works best for large listed firms that post consistently.
Vacancy duration as a measure of labor market tightness
Posting level records with dates give researchers the time between a vacancy appearing and being removed, which serves as a firm side measure of how hard a position is to fill and has been used to study everything from future profitability to whether workers avoid employers with poor environmental records after a local disaster. The measure depends on deduplication and on how removal is detected: a posting pulled because the requisition was cancelled looks identical to one pulled because it was filled, and firms that refresh listings on a schedule generate spurious short durations. Comparing durations across two independently collected posting panels is the usual check on whether a duration result is a labor market fact or a collection artifact.
Job posting datasets available on Dewey
LinkUp
Job postings collected directly from employer websites with posting and removal dates, which gives labor economists firm level vacancy counts and vacancy duration without the duplication that comes from aggregator feeds.
View datasetLightcast
Labor market and workforce data with occupation, skill and credential coding applied to postings, useful for building skill demand measures that map onto standard occupational classifications.
View datasetWageScape
Salary and compensation data drawn from advertised pay, letting researchers study posted wages for the same vacancies rather than inferring pay from realised earnings in survey data.
View datasetMore published work using job posting data
Faezeh Khosravi, Elaine M. Liu · working paper · 2026
Feng Dong, John A. Doukas, Rongyao Gloria Zhang · The Journal of Financial Research · 2025
Yongoh Roh · The Accounting Review · 2025
.svg)