The Occupation Pivot Shortlist, Ranked From Public Labor Data

You have a wishlist of roles you could pivot into and no reliable way to tell which ones will still be hiring after you have paid for the retraining. This guide is for a career changer who needs to narrow that list to a defensible few before spending money and months. It delivers a repeatable method: pull free public labor data, score each candidate occupation, and produce a ranked shortlist of three to five targets you can commit to.

Other guides in this library count real openings for the role you already hold, or score a single metro, or triangulate pay for one target. None of them help you decide which new occupation to aim at. That is the gap this document fills, and it does so while stopping you from chasing a headline growth rate that hides a tiny, low-opening field.

What "growing and hiring" actually means in the data

An occupation is worth retraining toward only if it will still be producing openings a decade from now, and openings come from three sources, not one. Growth is only the visible part. The larger part, usually, is replacement demand: people transferring out to another occupation and people leaving the labor force.

Total openings is the sum of occupational growth, occupational transfers, and labor-force exits. Transfers and exits together are called separations, and they are the primary driver of job openings. This is the single most important idea in the whole method, because it explains why "which occupations are growing" is the wrong question to ask first. A large, slow-growing field can out-produce a small, fast-growing one on sheer volume of annual openings.

That accounting example is the pattern in miniature. Jobs increased by only 500, but total openings were more than double at 1,100 because separations added another 600. If you had ranked that field on net growth alone, you would have understated its actual hiring by more than half.

There is also a methodology break you need to know about, because it makes older career advice wrong. The separations method was first used with the 2016-26 projections, replacing the cohort-component "replacements" method that BLS used from 1991 through 2014-24. The new method counts transfers to other occupation groups, not just deaths and retirements, which sharply raised measured replacement demand for older, high-exit occupations. Any guide written before that change understates openings for exactly the stable fields a pivoter should be considering.

The three free sources, and what each one covers

Three interlocking federal sources give you everything the method needs, at no cost. BLS Employment Projections is the national spine; Projections Central localizes it to your state; CareerOneStop packages both plus wages by geography. You will touch all three.

BLS Employment Projections publishes more than 800 detailed occupations nationally on a ten-year horizon, released roughly every two years. The 2024-34 vintage is live across the Occupational Outlook Handbook and CareerOneStop, with a 2025-35 release noted as upcoming. For scale: the economy is projected to add 6.7 million jobs from 2023 to 2033, reaching total employment of 174.6 million, growing 0.4 percent annually - notably slower than the 1.3 percent annual growth recorded over 2013 to 2023.

Projections Central hosts the state-level data. For hundreds of detailed occupations it provides initial-year employment, projected employment a decade later, the numeric difference, the percentage change, and average annual openings. It offers both long-term projections for 2024 to 2034 and short-term projections for 2025 to 2027, downloadable directly. CareerOneStop's Occupation Profile covers more than 900 careers, including outlook, projected growth or decline over ten years nationally and by state, and typical wages, with wages drawn from May 2024 estimates.

The common thread is that every state agency builds its projections from the same classification and wage backbone. Each State Employment Security Agency, working with BLS, uses the OEWS report to gather occupational employment data; those OEWS data feed the staffing patterns behind the projections, and everything reflects the SOC. Because the join key is identical everywhere, you can move a single occupation cleanly from national to state to metro without translation.

Why base size beats growth rate

The most common mistake in career-pivot research is ranking targets on percent growth, and it is a structural error, not a matter of taste. A high percentage on a small base produces few actual jobs. The fix is to require an absolute annual-openings floor before growth rate is allowed to matter at all.

The canonical case is documented and worth memorizing. Over one projection period, physician assistants were projected to grow roughly 48 percent, about four times faster than elementary school teachers at roughly 12 percent. Yet elementary teachers had about six times more new jobs, because the teacher base was around 1.8 million against a PA base near 66,000. Four times the rate, one sixth the jobs.

A shortlist that ranks on percent growth systematically over-weights small fields you will struggle to find a job in.

This is why the method leads with openings and treats growth as a tiebreak. Percent change tells you the trajectory; absolute openings tell you how many doors will actually be open when you finish retraining. Both matter, but in that order.

The scoring matrix: bands, not decimals

Score each occupation on two axes at once - how many openings it produces and how fast it is growing - and read off a single label, because BLS itself warns that projections are directional, not precise. Focusing on the direction and relative size of projected changes yields similar insight to chasing exact values, and it protects you from false precision.

BLS benchmarks each occupation against the all-occupation average rather than publishing fixed numeric cut points. The average projected growth for all occupations from 2024 to 2034 is 3 percent, and BLS uses relative descriptors like "much faster than average" down through "decline" anchored to that number. The exact percentage boundaries of the bands are not established publicly in current documentation, so do not invent them; use the 3 percent average as your dividing line and band above and below it.

The cleanest banded public system is New York's, and it is the model to copy. New York assigns four employment-prospect descriptors from two inputs: total openings and growth rate. Openings are cut into terciles - Small is the 25 percent of occupations with the fewest openings, Medium is the middle 50 percent, and Large is the 25 percent with the most. Growth is sorted into bands relative to the average. A moderate-growth occupation with large openings, or a fast-growing one with medium or large openings, earns "very favorable." Small or medium openings with declining growth earns "very unfavorable."

The two "target" zones matter most. The top-right, growing and large, is obvious. The bottom-right, the high-churn workhorse, is where the separations insight pays off: a flat or slow-growing field with a huge base still throws off enormous replacement demand every year, and pivoters routinely dismiss it because the growth headline looks dull.

Run the shortlist: the procedure

Work each candidate occupation through these seven stages in order. The whole pass takes most of a working day for a wishlist of ten to fifteen titles, and the output is a ranked shortlist of three to five targets with a stated reason on each.

A worked row makes it concrete. Suppose "data scientist" is on your wishlist. Step one pins it to its SOC. Step two records national growth, average annual openings, and the median wage. Step three tells you what share of those openings is replacement versus new jobs. Step four re-pulls all of that for your state and metro, because a nationally hot occupation can be flat where you live. Steps five and six place it on the grid and rank it against your other candidates. Step seven confirms the openings are not a rounding artifact on a small base.

For a quick supply read alongside the projections, a talent index shows where a role is thin relative to demand faster than biennial data can. In Refolk's index, US data scientist profiles number 28,714 against 5,714 in the UK, a 5.0x gap - the kind of cross-market signal that surfaces before the next projection round updates. If you want to check the supply and pivot path for a specific target, Refolk can pull real people who have already made the move.

Where this goes wrong

Seven failure modes account for nearly every bad shortlist, and most of them produce a target that looks great on paper and hires almost no one. Treat this section as the checklist you run against your own draft before you trust it.

The dominant failure is percent-growth headline chasing. A high rate on a tiny base yields few real openings, and the elementary-teacher versus PA case is the proof: a 12 percent field beat a 48 percent field on new jobs by roughly six to one. When it lies, it looks like a shiny "much faster than average" badge on an occupation with a few hundred annual openings nationwide. The check is a hard absolute-openings floor.

The second is ignoring replacement demand. A flat or declining occupation can still throw off large openings from exits and transfers, and a shortlist that reads only the growth column will wrongly dismiss a stable, high-churn field. The check is to read the separations split, not just the headline growth.

The remaining five are quieter but just as damaging.

On the wrong-SOC point specifically: mapping a title to an "All Other" code that ends in 9099 pulls a catch-all record, and some state agencies, such as New Jersey, exclude 99-ending "All Other" codes from demand lists entirely. That means the exact record you are ranking might be one the state does not even publish demand for. The Direct Match Title File exists to prevent this - each direct-match title maps to a single detailed SOC occupation, so all workers with that title fall into one code.

On the vintage point: state projections are prepared every two years and are not updated between publication years. If you pull national figures from the current round and state figures from an older one, your local rank is corrupted in ways you cannot see. Write the vintage next to every number.

Reading supply thinness inside a crowded field

When two occupations score similarly on the matrix, the tiebreaker that public projections cannot give you is how crowded each field already is - and a talent index can. Within healthcare, the advanced-practice tier is far scarcer than the base tier it grows out of, which is exactly the kind of headroom a pivoter wants to see.

That ratio is a leading signal. It says the pool of people who could pivot up into the scarcer, better-paid tier is enormous relative to the number who have. Projections tell you the NP role is growing; the supply ratio tells you it is not yet crowded with the very people best positioned to enter it. The wage gap reinforces the point: the life, physical, and social science group had a median wage of $78,980 in May 2024 against $49,500 for all occupations, a reminder that scarcer tiers usually pay more.

Use this as a second lens, never the primary one. Supply thinness is a signal about competition, not about whether the occupation hires; the projections still decide that. But between two matrix-equal finalists, prefer the one where the scarcer tier sits above a large, obvious feeder pool - that is where a determined pivoter has the clearest runway. Searching Refolk for people who have already made a given move, such as registered nurses who became nurse practitioners in a specific state, shows both the path and the employers hiring at the far end.

Keep the shortlist current

A pivot shortlist is a snapshot of a modeled decade, so it needs a maintenance rhythm rather than a one-time verdict. The projections update on a biennial cadence, and your local market moves faster than that. Two habits keep the work honest.

First, re-run the localization step when a new vintage lands. BLS Employment Projections refreshes roughly every two years, with a 2025-35 release noted as upcoming, and state files follow. When they do, re-pull every finalist and confirm the ranks still hold; a role that was Large-openings in one round can slip. Second, cross-reference live postings continuously, because projected openings are a modeled annual average, not a count of vacancies open today. A finalist that models beautifully but has no current postings in your metro is a warning, not a target.

The output you keep is small on purpose: three to five occupations, each with a band label, an openings figure, a wage, and a sentence explaining why it survived. That document is what you point your retraining budget at, and it is far more defensible than a list of roles that merely sounded like they were growing.