The Role Market-Tightness Read: Tight, Balanced, or Loose

A hiring manager asks the question that decides a plan: how hard will this role be to fill here? This guide is for workforce planners, talent-intelligence analysts, and strategy teams who need to answer that with a number, not an opinion. It takes you from a raw role-and-metro definition to a documented supply-to-demand ratio, a tight / balanced / loose verdict, and an error band, using only public data, so you can defend every link in the chain.

The deliverable is one sentence a manager can act on: "For this role in this metro, there are roughly X open requisitions per qualified, reachable candidate, which is tighter / looser than the national market, plus or minus Y." Everything below exists to make that sentence defensible.

What tightness means and the one number to anchor on

Labor-market tightness is the ratio of vacancies to jobseekers, written theta = v/u: the number of vacant jobs per available worker. It has a hard numeric anchor that most talent discussions lack. The efficient, balanced point is exactly one. Above one the market is inefficiently tight; below one it is inefficiently slack. As a real reference, the US had roughly twice as many vacancies as unemployed workers in March 2022, a theta near 2.0 and a documented "very tight" reading.

That anchor is why this method ends in a briefable number rather than a definition. You are not inventing a scale. You are mapping your role-metro figure onto an economic threshold that already exists.

There is a refinement worth knowing. The raw denominator, unemployment, undercounts who is actually available for a role, because most qualified people for a given job already have one. A worker-availability measure replaces raw unemployment with effective searchers, a weighted sum: ES = Us + 0.48·Ul + 0.4·Z_want + 0.09·Z_do-not-want + 0.07·N, where Us is short-term unemployed and Ul is long-term. Generalized tightness goes further, weighting the unemployed, the out-of-labor-force, and the already-employed by their relative search intensities. You will apply the same logic at the role level in the supply discount below, so hold that thought: supply is never just the unemployed.

Why the demand side is the weak half

The structural fact to internalize before you pull a single number: metro demand data is weaker than metro supply data, so your error band will almost always be wider on demand. Plan accordingly.

The federal benchmark for openings is BLS JOLTS. It measures job openings as a stock on the last business day of the month, counting a position as open if the job exists, work is available, it can start within 30 days, and it is being actively recruited. That is a clean definition. The problem is granularity. The JOLTS sample of about 16,000 establishments does not directly support state or sub-state estimates. Experimental metro estimates exist only for the 18 largest metros with 1.5 million-plus employees, built by combining sample with model-based estimates and smoothed by a three-month moving average. And JOLTS has no occupation detail at all.

So for a role in a metro, JOLTS cannot help you. You fall back to job-posting aggregators that scrape large numbers of sites, and those carry duplication and scraping error you have to remove by hand. Supply, by contrast, is directly countable from public profiles. That asymmetry is the first thing to tell a reviewer.

The discount chain: raw counts are not supply or demand

Both sides of the ratio start as inflated raw counts, and both need a documented discount before they can be divided. Demand inflates because one job appears on many boards and gets reposted. Supply inflates because most profiles are not qualified, reachable, or comp-fit. The table below is the toolkit; apply each lever as a sequential multiplier.

Each lever proves something, and each can lie. Read them that way.

Posting intensity. This is the ratio of total postings to unique postings. It proves how aggressively a market reposts. It lies when you skip it: a single ad reposted daily can count about six times a year under a 60-day window, so a market that looks tight may be one employer with a loud posting habit. Up to 80% of collected postings deduplicate away. The 60-day rule treats the first appearance as the original, removes duplicates for 60 days, then counts a still-live ad as new, starting a fresh window.

Qualified-applicant rate. This is the share of applicants who advance from application to screening. It proves that raw supply overstates hireable supply, and it does so unevenly. It was 11.5% overall but only about 5% in software and tech. The same raw pool size converts very differently by field, so you cannot carry one rate across roles.

Active share. About 30% of a talent pool is actively searching; roughly 70% is passive. This proves that accessibility, not headcount, caps near-term supply. It lies if you count every profile as reachable. But it also lies the other way if you count only the unemployed, because passive people can be moved. The honest move is to count active plus a reachable slice of passive, and state the slice.

Credential gating. Healthcare needs about 47 applicants per hire versus 191 in tech, because licenses pre-filter the pool. This proves that the funnel shape, and therefore the ratio, is role-specific before you touch any location data.

Counting supply: what the index actually shows

Supply is the countable half. The point of the method is to start from a real, directly counted number and discount it on the record, rather than guessing a pool size. The table below is drawn from Refolk's index of professional profiles and shows why raw counts need both a geography and a seniority read before they mean anything.

Two readings fall straight out of this. First, geography multiplies scarcity more than seniority does. The US React pool is 13.0x Germany's, a gap no amount of relaxing the seniority bar inside one country will close. Second, raising the bar costs less than it feels like: senior engineers still hold 68.6% of the US React pool. If the manager's instinct is "we will just hire junior and train," the data says the junior pool is the smaller one here.

Now discount it. Take the 5,449 US React engineers. Apply the active share (about 30%) to get the reachable-plus-active slice, then the software/tech qualified rate (about 5%) for the share who would clear screening. The number collapses fast, which is the point: the tech funnel is punishingly narrow, so a raw pool of 5,449 behaves nothing like a generalist pool of the same size. This is where a countable starting point earns its keep. If you had guessed the pool, you could not defend the discount.

Refolk is where I count the supply side in plain English: you ask for the role and place you are sizing and get the matching profiles back, which is the raw number you then discount. For the demand side you still go to postings, because that is where openings live.

Run the procedure

This is the full method, start to finish. Budget about a day of analyst time. Each step ends with a concrete "done" so you know when to move on.

On step one, a note that saves rework: anchor on a standardized occupation where you can, not raw title. A title-only match will treat near-identical roles as distinct, or collapse distinct ones. On step six, keep units straight: both your role-metro figure and the national benchmark must read as openings per available candidate, or the comparison is meaningless.

How this goes wrong

Most bad tightness reads fail in one of seven predictable ways. This is the section to keep open while you work, because a number that survives these checks is briefable.

Three of these deserve more than a row.

Over-collapsing distinct roles. A match on title and company alone will collapse genuinely distinct roles in different cities. Your de-duplication must key on location as well, or you will quietly delete real demand and call the market looser than it is. This is the mirror image of counting reposts, and both corrupt the demand number in opposite directions.

Ignoring passive supply. Supply built only on the unemployed overstates tightness, because the people qualified for a professional role mostly already have one. Generalized tightness exists precisely because the employed and the out-of-labor-force belong in the count, weighted by how hard they search. Whatever slice of passive candidates you decide is reachable, write it down. The verdict can hinge on it.

Point estimate with no band. Published funnel rates disagree by an order of magnitude depending on how the numerator and denominator are defined. A single number pretends that disagreement away. The fix is mechanical: carry the low and high of each discount rate through the chain and report the range. A reviewer who distrusts your 5% qualified rate can see what happens at 8% without redoing your work.

A tightness number without an error band is a guess wearing the costume of a measurement. </pull>

Reading the ratio and benchmarking it

Once you have discounted demand over discounted supply, you have openings per qualified candidate. Map it to the same scale tightness has always used. Below one is loose. Around one is balanced. Around two is very tight. The band labels come from the theta = 1 efficiency criterion, and the 2.0 reference is the documented March 2022 datapoint.

The matrix below is the judgement call to make before you deliver, because a headline number means different things depending on how much you trust each side.

Finish with the benchmark. State your role-metro ratio next to national vacancies-per-jobseeker, in the same units, and say which is tighter. "This role in this metro runs about 1.7 openings per qualified candidate, against a national market near balance, so it is meaningfully tighter than the economy as a whole." That single comparison is what turns a local number into a decision input: it tells the manager whether the difficulty is this role, this place, or just the market everyone faces.

Keep the read current

A tightness read has a shelf life. Postings are a flow that moves with hiring cycles, and the discount rates themselves get re-benchmarked. Rather than memorizing a value, re-run the mechanism. Re-pull demand on the same date-window logic, re-count supply from the index, and re-check whether the qualified-applicant and active-share rates you borrowed have moved before you reuse them. The verdict can flip without any change in the underlying pool simply because reposting intensity spiked.

Before you send the number, run the checklist.

The habit that makes this durable is refusing to hand over a number you cannot walk backward. If someone questions the verdict, you should be able to name the raw supply count, the three discounts, the raw demand count, the posting intensity, and the national benchmark, in order, and show where the range comes from. That is the difference between a tightness read a planner bookmarks and a figure that evaporates under the first hard question.