Selling the Job:

Pay Transparency and Advertised Pay in Fortune 500 Job Opportunities

March 2025 Spotlight

Imagine a job at a Fortune 500 company – perhaps you picture a finance position in New York City, or a tech job at Google or Amazon. But what do Fortune 500 job opportunities actually look like? This series from the National Labor Exchange (NLx) Research Hub utilizes our job feed data to provide insights to those supporting the workforce system, from case managers or business service representatives at a local American Job Center, to policy analysts or administrators at a central office, recruiters, and everyone in between. If you are trying to connect workers to good jobs, looking at jobs with financially stable companies is critical, and where better to start than with large high-revenue companies.

Pay transparency for prospective employees is becoming increasingly common, with more cities, counties, and states passing pay transparency laws every year. Pay transparency broadly refers to the employer practice of disclosing wage or salary information, with the level of transparency often influenced by state and local mandates or employer pay strategy. Pay transparency for prospective employees means that employers provide pay information (usually in the form of expected pay ranges) to applicants in the job posting, upon request, or at the time of final interview, for example. What does this type of pay transparency look like in practice? How transparent are job postings, and what do advertised wages look like when they are included?

1 Throughout this analysis, it is important to keep in mind that Fortune 500 employers differ from other employers in myriad ways that could affect how they handle pay transparency, so the findings here don’t necessarily reflect pay transparency for other job opportunities. One example is that Fortune 500 employers are more likely to list job opportunities across state lines. They might also have larger recruitment budgets and more standardized job listing procedures, including for setting and advertising pay levels in accordance with changes in state and local laws.

Overall, most Fortune 500 job postings in 2023 were not pay transparent.

2 Note that this analysis does not delve into why certain occupations are more pay transparent than others. It is possible that not one single explanation applies to each occupational group. For example, it is possible that certain occupations are more likely to be covered by laws regarding pay transparency than others (either by geographic location or employer type). It is also possible that different job market conditions cause employers to utilize different recruitment strategies that affect pay information in job postings for some occupations and not others.

Fortune 500 job opportunities in states with pay transparency laws were more likely to include pay information.

3 NY’s statewide law went into effect in September of 2023, but there were already a number of existing local laws in areas such as New York City, Ithaca, and Westchester County, so we counted NY as a “pay transparent” state for this analysis.

4 We only present two odds ratios from our logistic regression here, with asterisks for statistical significance. Please see the Methodology section for more detail on the regression and all factors that were included in the model.

5 The classification of below-average paying occupation is based on actual average pay for jobs across all employers (beyond just Fortune 500s).

6 When included, pay was reported just over half the time in hourly rates (55%) and just under half the time in annual rates (43%). About two percent of postings used other units such as weekly or monthly rates.

7 $16-22/hour has a $6 range and $19 midpoint which equals a range that is 33% of the midpoint. $125,000-$175,000/year has a $50,000 range and $150,000 midpoint which also equals a range that is 33% of the midpoint.

8 Postings advertising exact pay points or pay ranges of 0 are excluded from this analysis.

9 We also considered how below average paying occupations might come into play. While it is true that job postings for below average paying occupations are more likely to be reported in hourly rates, the driving factor for differences in advertised pay ranges remains the pay unit. Accounting for other factors such as whether the job opportunity is for a below average paying occupation does not alter the pay preciseness by much (hourly becomes 18% of the midpoint and annual becomes 37% of the midpoint). Please see the Methodology section for more detail.

10 Neither potential explanation is inherently positive or negative, however, as long as the employer adheres to best practices for fair pay when coming to an agreed upon wage or salary with the employee.

• About half of Fortune 500 job postings overall had pay range midpoints below and half had pay range midpoints above the actual national median pay (46% below and 52% above).

• Over half (57%) of Fortune 500 job postings reporting pay information in hourly rates had pay range midpoints below the actual national median pay. The median pay range midpoint for those job postings was $18.50/hour compared to the actual national median of $18.71/hour.

• Two thirds (66%) of Fortune 500 job postings reporting pay information in annual rates had pay range midpoints above the actual national median pay. The median pay range midpoint for those job postings was $118,750/year compared to the actual national median of $99,580/year.

11 When included, pay was reported just over half the time in hourly rates (55%) and just under half the time in annual rates (43%). About two percent of postings used other units such as weekly or monthly rates.

12 Note that we were not able to match 2023 wages from Fortune 500 employers specifically to their advertised wages. Instead, we used national wages and compared those to Fortune 500 advertised pay by occupation.

METHODOLOGY

Sample

We utilized the same 2023 Fortune 500 dataset from the previous Spotlights with 5.5 million postings, or over a quarter of postings in the NLx in 2023. Please review the methodology section from the first Spotlight in this series for information on how the dataset was compiled and information about the sample.

Identifying pay information

We utilized a version of the Job Posting Structure (JPS) Parser, which makes use of Claude/Anthropic GenAI to process and parse out information from job descriptions. The Research Hub is still in the process of developing the JPS Parser, so past and future versions might differ slightly from the one used for this analysis. The original extract prompts are publicly available on the Research Hub GitHub. The version used for this analysis specifically asked for the pay range minimum, pay range maximum, and pay range unit (hourly, annual, etc.).

We randomly selected 2,000 job postings in the dataset to manually check parsed-out pay information against the original job description. The results of our check are below:

• 95% of sampled job postings were accurate in terms of identifying whether or not pay info was present in the description. The vast majority of the ones the accurately identified when pay information was present (99%) provided correct information. Some descriptions included multiple pay ranges for different geographic locations, for example. Those are counted as correct as long as the prompt extracted at least one of the ranges or took the absolute minimum and maximum across all ranges. The Research Hub is planning to adjust future versions of the JPS Parser to account for these situations.

  • Note that some job descriptions included links to other locations with salary schedules. In addition to being difficult for the JPS Parser to extract, this pay information is not strictly provided in the job posting itself, so these types of postings are not counted as pay transparent in this analysis.

• 5% of sampled job postings were inaccurate in terms of identifying whether or not pay information was present in the description.

  • Most of these (70%) were false negatives where the model didn’t know what to do when there were multiple pay ranges or the description included a min-median-max format instead of just the min and max. The Research Hub is planning to adjust future versions of the JPS Parser to account for these situations.

  • Some of these (18%) were true false positives where the model pulled something such as number of employees or hours per week instead of pay information.

  • A small amount of these (12%) were flagged as false positives only because the model added 0s to the pay min and max fields when there was no pay information present in the description (not to be confused with the rare cases where the description actually did have an advertised pay range of $0.00-$0.00). Either way, zeroes were re-coded as missing for this analysis.

Overall, we estimate a 95% accuracy rate of the pay information used in this analysis across all error types and after re-coding zeroes.

Analysis

The analysis in this Spotlight is based on simple medians and ranges (reported pay maximum minus reported pay minimum). Generally, we utilized medians instead of means to avoid undue influence from outliers that might either occur within the job posting or from an error in the parsing process. To look at factors influencing pay transparency, we ran a logistic regression predicting pay transparency by occupation group, industry, if the posting was in a pay transparent state, and if the posting was for a below average paying occupation, and clustering by employer. In looking at advertised pay ranges, we ran a quantile regression predicting the median range as a percentage of the midpoint by pay unit, occupation group, industry, if the posting was in a pay transparent state, and if the posting was for a below average paying occupation, and clustering by employer. To compare advertised pay ranges to actual national pay by occupation, we incorporated external data on median pay by occupation (reported both in hourly and annual rates) obtained from the Bureau of Labor Statistics (BLS) Occupation Employment and Wage Statistics tables and merged it with individual job postings.

Author: Marissa Hashizume, National Labor Exchange Research Hub