A Job Posting Is an Ad: What Postings Can and Can’t Tell Program Planners

When a college asks whether its region needs another program, the fastest evidence on the table is often a count of online job postings. It feels current and specific. It is also a record of advertising, which answers some planning questions well and others badly.

One report, read closely

In January 2025, the Orange County Center of Excellence for Labor Market Research evaluated a proposed general electrician program for the Los Angeles and Orange County region. Its report counted 1,817 online job postings for electricians over the previous 12 months, and the largest single poster in the region was Aerotek, with 102 postings, or 6% of the total.[1]Aerotek describes itself as a staffing agency,[2] so the name at the top of the list is a recruiter working on behalf of other companies, not necessarily the place an electrician would work.

The center did not rest its recommendation on those postings. It compared 2,349 projected annual openings for electricians across the two counties with an average of 815 awards a year from educational institutions. It weighed the occupation’s typical entry-level education and its entry-level wages against a local living-wage standard, and it endorsed the program because some of its criteria were met.[1]It also printed a caveat anyone reading a posting dashboard should keep in view: “A single job posting may not represent a single job opening.”[1]

The report’s postings came from Lightcast, a labor market analytics firm.[1]Lightcast’s own methodology makes the same point from the provider’s side: a posting is not necessarily a vacancy, and posting counts are better understood as a measure of employers’ recruitment marketing.[3]

Four numbers that are easy to confuse

A posting, an opening, a hire, and a projected opening are different measurements, collected differently, answering different questions.

• A job opening, in the federal Job Openings and Labor Turnover Survey (JOLTS), is a specific position with work available that could start within 30 days and for which the employer is actively recruiting workers from outside that location. Openings are counted on the last business day of the month.[4]
• A hire is any addition to the payroll during the month.[4]
• A posting is an advertisement collected from websites and counted by a data provider according to its own rules for removing duplicates and retiring old ads.[3]
• A projected occupational opening, in the Bureau of Labor Statistics (BLS) projections, estimates how many workers an occupation will need each year to replace people who leave the labor force or move to other occupations, plus any growth.[7]

The scale differs accordingly. In 2025, U.S. employers averaged 7.1 million job openings and made 63.0 million hires.[5]BLS projects about 17.5 million occupational openings a year for 2025 to 2035.[6]These figures do not contradict one another, and they should not be turned into ratios. The 7.1 million is an average of month-end snapshots, the 63.0 million hires are a full year’s flow, and the 17.5 million is a projection that leaves out workers who change employers but stay in the same occupation.[7]BLS also cautions that its projected openings “are not direct estimates of training needs for an occupation.”[8]

Who shows up online, and who doesn’t

For community colleges, the most consequential limitation is coverage. Postings capture some parts of the labor market far better than others, and the gaps are not random.

The most recent evidence we found comes from a September 2026 National Bureau of Economic Research working paper by Michael Dalton of BLS, Lisa Kahn, and Andreas Mueller. It links job-posting data from Burning Glass, a data company whose postings are now collected by Lightcast, to confidential JOLTS survey records for the same establishments.[10]Coverage has improved sharply. To match openings measured from JOLTS, the researchers had to multiply posting counts by about 30 in 2007, about 10 by 2011, and about 4 in later years.[10]

Representation remains uneven. In the 2011–2019 data the authors use for these comparisons, postings from the lowest-paying fifth of establishments had to be multiplied by 17 to match JOLTS openings, compared with 2.8 for the highest-paying fifth. Postings in large central metro areas needed a factor of about 4.5; those in the most rural “noncore” areas needed about 16. Construction and leisure and hospitality required large adjustments, while professional services required much smaller ones.[10]

That unevenness has limits. Because large establishments account for most openings and are well represented online, the authors find that postings still capture the overall distribution of openings well, and that measures of employer skill demand change only modestly after adjustment.[10]

The same study found that about 40% of postings came from establishments in months with no corresponding opening in the JOLTS data. The authors attribute that mismatch to measurement error or to postings that fall outside JOLTS definitions of openings and hires, and they show it is confined to small firms; for larger firms, the gap largely disappears.[10]The paper is new and has not yet been peer reviewed.[10]

Older evidence speaks to education. In a 2014 technical report built on 2013 data, Georgetown University’s Center on Education and the Workforce estimated that 80% to 90% of openings requiring at least a bachelor’s degree appeared online, compared with 30% to 40% of openings for candidates with some college or an associate degree.[9]The authors described the estimate as back-of-the-envelope, and coverage has improved since then.[9][10]

We found no newer U.S. estimate of how much of each education level’s openings appear online. The 2026 study offers a partial check: after reweighting postings to match JOLTS, the share of postings specifying an education requirement fell by about 6 percentage points from 56%, a correction the authors call modest.[10]

A count is something someone builds

Every posting total reflects decisions about what to count. Lightcast deduplicates postings across sources within a 60-day window and expires an ad after no more than 120 days; by its own example, the same ad posted every day on different sites for a year would be counted about six times.[3]The National Labor Exchange duplicates remote and multi-location postings into each state or location where hiring might occur, and its own researchers note that this overcounts at the state level.[11]The Conference Board’s Help Wanted OnLine series excludes aggregator job boards and targets staffing-firm ads for removal.[12]

Changes in the posting business can move the numbers, too. Between the end of 2012 and the end of 2015, JOLTS openings rose 48% while a composite index of online postings rose 4%. Federal Reserve economists estimated that Craigslist’s price changes for job ads could by themselves explain roughly a quarter of the gap.[13]

None of this makes postings unusable. It means a posting total is partly a product specification. Before comparing two dashboards, or the same dashboard across years, it is worth asking the provider how it treats duplicates, reposts, staffing ads, and jobs listed in several locations.

What an ad asks for is not who gets hired

Posting language is often used to decide which credential a program should lead to. The evidence counsels caution. The Burning Glass Institute, a research organization, and Harvard Business School’s Project on Managing the Future of Work followed hiring into about 11,300 roles at large firms where hiring could be observed before and after the firm removed a bachelor’s-degree requirement from its postings. On average, the share of people hired without a degree rose by about 3.5 percentage points, and about 45% of firms appeared to make the change “in name only,” with no meaningful difference in actual hiring.[14]

Across the economy, the authors estimated the net effect at 0.14 percentage points, “not even 1 in 700 hires.”[14]The study concerns bachelor’s degrees at large firms, but the caution travels: a requirement added to or dropped from an ad is not proof that hiring changed.

Requirements also move with the labor market. During the Great Recession, when job seekers were plentiful, employers asked for more: economists Modestino, Shoag, and Ballance estimate that the rise in unemployed workers could account for 18% to 25% of the increase in skill requirements between 2007 and 2010.[16]As unemployment fell between 2010 and 2014, the share of postings requiring at least a bachelor’s degree fell with it, by roughly 0.27 percentage points for each one-point drop in local unemployment.[15]A requirement in this year’s postings may say as much about how easy it is to hire as about the work itself.

Where postings earn their place

Postings still capture several things that are hard to see anywhere else, and they capture them quickly.

Skill and tool language. Postings are a detailed public record of what employers say the work involves, and that detail varies from employer to employer within the same occupation. In one study of postings for professional jobs, ten skill measures accounted for roughly 15% of the variation in wages across firms.[17]The same authors caution that employers leave out skills they take for granted and sometimes list requirements to signal something about the job.[17]For a curriculum committee, postings are a list of claims to check with employers, not a specification.

Emerging skills. Postings register new technology as it arrives: from 2010 to 2018, AI-related vacancies grew faster in establishments whose work was more exposed to AI.[18]They can also keep a trend in proportion. In 2025, 2.6% of U.S. postings required AI skills, according to Lightcast data compiled for Stanford’s AI Index.[19]

Timing. Postings can register a turn within weeks. At the start of the pandemic, vacancy postings collapsed in the second half of March 2020 and were down more than 40% by late April, a pattern that unemployment claims and BLS employment data largely matched.[20]

Employer identity. A posting names who is recruiting, whether that is the employer or a staffing firm working for one, as in the electrician report. That can open an employer conversation. It cannot settle how concentrated local hiring is, meaning whether a few employers account for most openings: the 2026 BLS-linked study found that concentration measures built from postings overstate the concentration of actual openings nearly threefold, because small and medium-sized establishments are underrepresented online.[10]

Advertised pay. Posted pay shows what employers are offering now; official wage estimates show what people already in the job were paid as of a past reference date. Many postings still omit it: in Indeed’s current data, 51.8% of U.S. postings listed explicit pay in August 2026, with lower shares in nursing (45.5%) and medical technician postings (37.5%).[21]

Advertised pay is also an offer range, not an outcome. In 2012–2017 postings, before pay-transparency laws spread, posted wages ran about 40% above BLS wage data in low-wage occupations and about 20% below it in high-wage occupations.[22]Official occupational wage estimates, from the BLS Occupational Employment and Wage Statistics (OEWS) program, measure something different: the straight-time pay of people already in the job, excluding premiums such as overtime and shift differentials, drawn from six survey panels collected over three years, with wages from earlier panels adjusted to the estimates’ reference date.[23]

About “ghost” postings

Postings for jobs that are not really open are a real concern, but no official statistic measures them. The Congressional Research Service notes that JOLTS openings are unlikely to include such postings if employers answer the survey honestly, and that the firms publishing prevalence estimates disclose varying amounts about their methods.[24]

In a December 2024 report, Greenhouse, a hiring-software company, said 18% to 22% of the jobs posted on its platform in a given quarter are classified as ghost jobs, without describing how it classifies them.[25]Nor should the 40% mismatch in the BLS-linked study be read as a ghost-job rate. As described above, its authors attribute it to measurement error or to postings that fall outside JOLTS definitions, concentrated in small firms.[10]

What the rules ask colleges to use

Federal career and technical education law points colleges toward labor market information and projections. Under Perkins V, each local needs assessment must describe how programs are aligned with in-demand industry sectors or occupations identified by state or local workforce development boards, or designed to meet local education or economic needs those boards have not identified. The statute’s ongoing stakeholder consultation may include ensuring programs of study are informed by labor market information and designed to meet current, intermediate, or long-term labor market projections.[26]The statute does not mention job postings.

California’s rules are more specific. Its 2026 program-approval handbook requires new career technical education program proposals to include labor market projections for the most applicable occupations and regions, where such data are available. It lists job advertisements in the college’s service area as one permissible form of supporting evidence, alongside employer surveys, industry studies, and employer letters, and again among the sources a college may use if labor market information is not available.[27]Projections are also what the Center of Excellence report in our opening led with.[1]

Matching the question to the evidence

No study we found tests whether programs designed with posting data achieve better completion, employment, or earnings outcomes. That is a gap in the evidence, not proof that posting data does not help. The practical lesson, then, rests on what the research shows about the data itself: ask postings only the questions they can answer, and pair them with evidence built for the rest.

Postings are among the fastest-moving evidence a program planner can see, and among the easiest to overread. They are strongest as a record of how employers describe and advertise work: the skills they name, the pay they offer where they disclose it, how quickly recruiting turns. They are weakest as a count of jobs. On the evidence above, they are probably weakest of all for lower-wage work, rural places, smaller employers, and, if the older estimates still hold, jobs below the bachelor’s level. Read them as ads, often informative ones, and let other evidence answer the questions ads were never built to answer.