What’s in the Data? (2/9)
Parsing Out Pay Information from Job Postings
Marissa Hashizume, NLx Research Hub Economist
February 9, 2026
Welcome to the blog series, What’s in the Data? These posts will provide updates on available features and interesting discoveries in the NLx Research Hub data. Whether you’re an analyst or researcher, or just curious about what job posting data looks like, this series will keep you informed. We encourage those interested in using the NLx Research Hub data to check out our data request page for more information.
Please note that job postings are not equivalent to job openings. These posts describe the NLx Research Hub job postings only.
Why do we need to parse out pay information? Advertised wage or salary information typically lives within chunks of text in the job description. In other words, if we want to use that information for analytic purposes, we need a way to find and extract it from the surrounding description text and do so at scale for the millions of records the NLx Research Hub gets every month. Fortunately, we live in a time of Large Language Models.
The NLx Research Hub partnered with Amazon Web Services utilizing their Nova Lite model to parse out pay and other information from our job posting data. Our initial testing produced 97% accuracy of Nova Lite’s pay extraction compared to manual coding. Here we describe what the pay data looks like for the 31.5 million NLx Research Hub job postings from 2024 and 2025.
Overall, 55% of job postings in 2024-2025 included pay information. By occupation group (2-digit SOC/ONET), pay transparency varied from 31% in Farming, Fishing, and Forestry occupations to 74% in Protective Service occupations and 72% in Computer and Math occupations.
Most Occupation Groups Had at Least 50% Pay Transparency
Pay transparency also varied by state and territory, with the Virgin Islands and Guam leading at 83% and 80%, respectively, closely followed by Colorado (78%), California (77%), Washington (75%), and New York (74%) – the four states that have had statewide pay transparency laws for job postings in place the longest.
Most U.S. States and Territories Had at Least 50% Pay Transparency
Pay information was advertised in hourly rates about half the time and in annual rates about half the time. However, most occupation groups were advertised predominately in one rate or the other. For example, Food Preparation and Serving occupations were advertised in hourly rates 95% of the time and Management occupations were advertised in annual rates 90% of the time.
Across occupation groups, advertised hourly wages typically ranged from $20-$25 and annual salaries typically ranged from $88k-$149k. Typical (median) midpoints were $22/hour and $120k/year. Among occupation groups predominately advertising pay in annual rates, Legal ($149k), Management ($147k), and Computer and Math ($143k) occupations topped out the list of salary midpoints. Among occupation groups predominately advertising pay in hourly rates, Healthcare Practitioner and Technical occupations had an exceptionally high median wage midpoint of $41.50. Installation, Maintenance, and Repair ($28.50) and Construction and Extraction ($27.50) occupations came next on the list of hourly wage midpoints.
Median Pay Midpoints by Occupation Group
Have we piqued your interest in pay information from job postings? We are looking for pilot users of our 2024-2025 parsed data! You can either 1) help us validate the accuracy of parsed information, 2) help us explore the microdata and make sure the information is in a useful format, 3) help us explore aggregate data and see how you might be able to use it in your work, or 4) all of the above. Please reach out to us at nlxresearchhub@naswa.org to get involved.
