Four Numbers Hidden Inside Every Jobs Projection
A labor-market projection can look like a single verdict—up, down, or growing fast—but the BLS occupational table is built from several different measurements. This chapter focuses on the relationship among employment level, percentage change, numeric change, and annual average openings. Learning what each number measures, and what it leaves out, lets a reader compare occupations without mistaking a rate for a hiring count or a net decline for a closed door.
The distinction is easiest to see when the columns are read as a sequence rather than as a ranking. First ask how large the occupation is. Then ask how many jobs separate the starting year from the ending year. Only after that should you ask how large that change is relative to the starting base. Finally, examine the annual flow of openings. The order matters because each step corrects a different possible misunderstanding.
Start with the stock: how large is the occupation?
The employment level is the stock of jobs in an occupation at a particular time. In the current BLS table, employment is reported in thousands for 2025 and projected 2035. The all-occupation row starts at 170,280.8 thousand jobs and reaches 176,198.5 thousand. That tells us the size of the labor market at the two endpoints, but it does not tell us how many people will be hired during the decade. Many of the positions counted at the end are the same positions that existed at the beginning, and people may move through them more than once.
Numeric employment change is the difference between those two levels. For all occupations, the projected change is 5,917.6 thousand jobs. This is a net change: the number of jobs in the ending year minus the number in the starting year. “Net” is doing important work here. If an occupation creates 100 positions and loses 40, its employment level rises by 60. The numeric change records the result, not every movement that produced it. It is the right measure for asking how much larger or smaller the occupation’s job stock is expected to become.
Percentage change puts that numeric difference against the starting level. In simplified form:
percentage change = numeric change ÷ starting employment × 100
For the all-occupation row, the 5,917.6-thousand increase is 3.5 percent of the 170,280.8-thousand starting base. The percentage makes it possible to compare categories of very different sizes. An increase of 60,000 jobs means something different in an occupation that began with 100,000 jobs than in one that began with 10 million. The rate places both changes on a common relative scale.
The time reference matters as well. The two employment levels are snapshots at the beginning and end of the projection, while the numeric change connects those snapshots. The percentage is a comparison of the change with the starting snapshot. None of these fields describes the path between the two dates. An occupation could grow quickly early in the decade and then level off, or grow slowly for years before expanding. The table summarizes the endpoints; it does not narrate every year in between. That is another reason not to treat the percentage as a promise about the next hiring cycle.
A useful table-reading habit is to say the unit aloud before interpreting the value. “170,280.8” in this table means thousands of jobs, not 170,280.8 individual openings. “3.5 percent” means the projected employment stock is larger by that proportion, not that 3.5 percent of workers will be newly hired. “5,917.6 thousand” means the ending stock exceeds the starting stock by that amount, not that exactly that many people will receive jobs. The language may feel repetitive, but it prevents three different quantities from collapsing into one vague idea of growth.
That normalization is useful, but it also hides the denominator. A 50 percent increase in a base of 20,000 jobs creates 10,000 additional jobs. A 5 percent increase in a base of 2 million creates 100,000. The second occupation grows more slowly by rate and much faster by number. Neither calculation is more “real.” They answer different questions: one identifies relative expansion, while the other identifies the size of the net addition.
The BLS industry projections provide a national example. Utilities leads the major-industry growth ranking at 9.8 percent, but the projected numeric increase is 58,800 jobs. Private healthcare and social assistance grows at 9.5 percent—slightly more slowly by rate—yet adds 2,204,800 jobs. The near-equal rates do not imply near-equal changes because the industries begin at different scales. Utilities’ rate may be a meaningful signal about electricity demand and investment. Healthcare’s numeric change is more informative for someone asking where millions of additional positions may exist.
The same logic applies to occupations, but the differences can become sharper because some occupations are small niches and others employ millions of people. A fast rate can identify an emerging specialty, a technology-related role, or a field responding to a focused demographic need. It does not establish that the field will have a large number of openings. A large numeric increase can indicate broad demand, but it does not establish that the new work will be easy to enter, well paid, or available in the reader’s region.
This is why it is helpful to keep two comparisons in view. Compare rates when the question is relative momentum: which category is expanding faster from its own base? Compare numeric changes when the question is scale: which category contributes more net jobs to the economy? If the question concerns a person’s chances of finding a position over several years, neither comparison is complete. That question requires the fourth number.
Read the flow: what do openings add?
Annual average occupational openings are not another way of saying net new jobs. They describe the average number of openings associated with an occupation over the projection period. Some arise because the occupation expands. Others arise because positions must be filled as workers leave, retire, or move into another occupational category. The average therefore describes a flow through the occupation, while employment level describes the stock of jobs at a point in time.
This distinction explains one of the most important counterexamples in the BLS table. Office and administrative support is projected to fall by 752.1 thousand jobs between 2025 and 2035, a 4.0 percent decline. Yet the group has 1,727.7 thousand annual-average openings. The projected ending stock is smaller, but the occupation is not frozen. Employers still need people to fill positions as the workforce changes. A reader who looks only at the negative percentage might conclude that there will be no entry points; a reader who looks only at the openings number might miss the fact that the long-run job stock is shrinking.
Production shows the same pattern on a smaller scale. It is projected to lose 33.5 thousand jobs, or 0.4 percent, while still producing 867.7 thousand annual-average openings. The openings do not cancel the decline, and the decline does not cancel the openings. They describe different dimensions of the same labor market: one describes the direction of the occupation’s size, and the other describes the expected volume of positions that become available over time.
Think of the employment level as the number of seats in a theater and openings as the number of times seats become available across many performances. If some audience members leave after one performance and others take their places, the theater can fill many seats over time without adding more seats. If the theater removes seats, the total capacity can decline even while many seats are still filled and refilled. The image is only a way to visualize the stock-and-flow relationship; the labor data are not measuring theater seats. The important point is that a flow can be large relative to a stock when movement through the stock is frequent.
This also explains why numeric change should not be used as a proxy for hiring demand. A shrinking occupation may be losing positions overall while still requiring employers to replace people who leave. A growing occupation may be adding positions but have fewer openings than its numeric increase suggests if its existing workforce is relatively stable. The relationship depends on both expansion and movement. The openings column gives readers information that the two endpoint employment levels cannot supply.
A large opening count can be generated by a large workforce with frequent turnover even when growth is modest. Fast food and counter workers are projected to grow 5.8 percent and add 223.3 thousand jobs, but the occupation has 827.4 thousand annual-average openings. Home health and personal care aides grow much faster, at 18.1 percent, and add 847.3 thousand jobs, with 760.5 thousand annual-average openings. Cashiers are projected to decline 6.5 percent and lose 200.6 thousand jobs, yet still have 521.3 thousand annual-average openings. These rows cannot be sorted into a simple “good” and “bad” ranking. They show high flow, different growth directions, and very different reasons a person might consider the work.
The word “average” also matters. An annual average smooths a decade of projected activity; it does not say that the same number of openings will appear every year. Economic conditions, employer decisions, retirements, local demand, and worker movement can vary. Nor is an annual-average opening necessarily a job advertisement visible to a job seeker today. It is a modeled estimate of opportunity volume, not a list of vacancies reserved for new entrants.
The Georgetown Center on Education and the Workforce makes another boundary explicit in its 2021–31 projections. Its annualized-openings metric counts new jobs and jobs created when workers move from one occupational category to another, but not job-to-job moves within the same occupational category. That definition is part of the meaning of the number. It also means readers should not silently treat CEW’s 18.5 million annualized openings as identical to the BLS table’s 17,461.8 thousand annual-average openings for 2025–35. The periods, data systems, and methods differ even though both sources are trying to describe labor-market opportunity over time.
Which number should receive the most attention depends on the decision. A student comparing fields may begin with numeric change and openings, then investigate preparation and wages. A career changer may care more about openings in the relevant occupation than about the sector’s percentage growth, because a high rate in a small occupation may offer few realistic entry points. An employer may use the employment base and projected change to estimate the scale of a future talent pool, then use openings to think about replacement and recruitment pressure. A policymaker or educator may need both: numeric change to anticipate expansion and openings to understand the continuing volume of people who must be prepared.
Suppose a career counselor is asked whether a student should pursue a rapidly growing specialty. The rate is a reasonable starting signal, but the counselor should then ask how many jobs the specialty contains, how many annual openings it produces, and what preparation those openings normally require. If the student instead needs work soon and can access a short training route, a large occupation with steady openings may deserve closer attention even if its projected rate is ordinary. If an employer is planning a new facility, numeric growth may matter more than replacement openings because the organization needs to estimate added capacity. The numbers do not dictate the decision; they reveal which question still needs evidence.
There is no universal “best” column. The mistake is to let the most eye-catching column answer a question it was never designed to answer. Percentage change is not a hiring forecast. Numeric change is not the number of people who will receive offers. Annual openings are not all newly created jobs, and they do not measure pay, retention, or accessibility. A projection becomes more informative as the measures are read together.
The four-number sequence also points to the next layer of interpretation. Once we know whether an occupation is large, fast-growing, expanding in absolute terms, or producing a large flow of openings, we still need to know what kind of work sits inside the category and how the category connects to the wider economy. A large opening count may reflect an industry’s scale, an occupation’s turnover, or both. Understanding that mechanism is what turns a table into a picture of how employment changes.