What Counts as a Tight Labor Market?
Local labor-market tightness is useful only if we are clear about what is being counted and whose experience the count is meant to represent. This chapter follows the measurement chain from a JOLTS-reported opening, through a state-level openings-per-person measure, to household reports of job changes and job quality. The goal is practical: to know which part of the offer-arrival mechanism a statistic captures, and which conclusions it cannot support.
That discipline matters because “job openings” can sound more direct than they are. An opening is not necessarily an offer, an online advertisement, a job that will remain available for a year, or a role that matches a particular worker’s skills and constraints. Likewise, a worker’s report that a new job is better is not an administrative measurement of a precise pay increase. The Federal Reserve study connects these different pieces, but it does not treat them as interchangeable.
From a filled position to a local tightness measure
The JOLTS definition begins with an establishment rather than an individual. A position counts as a job opening only when three conditions are met. There must be a specific position and work available for it. The job must be capable of starting within 30 days, whether or not the employer can find a suitable candidate in that time. And the establishment must be actively recruiting workers from outside the establishment.
That last condition separates an opening from a position reserved for an internal transfer or promotion. Positions with start dates more than 30 days away are excluded, as are roles for which someone has already been hired but has not yet reported. The measure is therefore intended to capture active external demand for labor at a particular moment, not every future staffing intention.
JOLTS also measures the opening at a specific time: the last business day of the reference month. Hires and separations are different. They count changes to the payroll during the entire month. This timing distinction is easy to overlook when the figures appear together in a release. A month-end opening is a stock of positions available at a point in time; a hire is a flow of additions over a period. Comparing them can be informative, but they do not describe the same event.
The Federal Reserve paper uses these openings to construct a local measure of tightness. Its preferred variable is the natural log of annual state-level openings per thousand residents. The authors average monthly JOLTS openings from November of the previous year through October of the survey year. That window matches the SHED survey’s retrospective questions about the preceding 12 months. The denominator is total state population rather than the labor force.
The denominator choice is not cosmetic. If the number of people in the labor force changes in response to labor-market conditions, dividing by the labor force can make the measure partly reflect the response it is supposed to explain. A population denominator is not perfect—it includes people who may not be looking for work—but it is less directly entangled with the participation decision. It answers a specific question: how many actively recruited positions are present relative to the size of the surrounding population?
The annual window also changes the meaning of the estimate. A worker interviewed in late 2024 is not matched to openings in one isolated month. They are matched to an average environment covering the period that could have influenced the job transitions reported in the survey. This reduces the temptation to link one unusual monthly spike to a retrospective job change, though it also means the measure is not a real-time index of the exact day a person made a decision.
The complete chain looks like this:
The household side of the chain comes from the Survey of Household Economics and Decisionmaking. The SHED is a nationally representative survey of about 11,000 adults conducted in October or November. The paper uses data from 2021 through 2024 and focuses its baseline analysis on prime-age adults, ages 25 to 54. Respondents report whether they worked for pay or profit in the prior month and whether, during the prior year, they applied for a new job, started a new job, quit, were laid off, asked for a raise, or received one.
Those questions measure transitions and reported actions, not the full history of a person’s search. A respondent who did not apply may have been satisfied, discouraged, constrained by care or geography, or unaware of suitable openings. The same “no” answer can arise from very different circumstances. The paper’s model gives those outcomes a mechanism, but the survey cannot observe every reason behind them.
The most distinctive questions go to people who were employed at the survey date, had started a new job during the prior 12 months, and reported changing their main job. They are asked whether the current job is better, the same, or worse than the one they had a year earlier. The comparisons cover pay or benefits, opportunities for advancement, interest in the work, physical demands, COVID-19 policies and exposure, and work-life balance, followed by an overall judgment of whether the new job is better.
That design has an important strength: the comparison is made by the same person across two jobs rather than by comparing the average wages of unrelated workers. It also has an important boundary. The result measures a perceived directional change among job changers. It does not tell us the dollar amount of a pay increase, the value every worker assigns to an amenity, or what happened to people who considered changing but stayed.
The way the paper codes the outcomes makes this boundary especially important. The indicator for an improved amenity is set to zero for people who were not eligible for the job-quality questions, including people who did not change jobs. That choice lets the estimate capture two margins at once: whether a person makes a move and whether the move improves the named characteristic. It is useful for the question “How does local tightness affect the chance that a prime-age person experiences an improvement?” It would be misleading to read the same coefficient as the average improvement among people who definitely changed jobs.
The distinction also explains why the paper reports both broad transition outcomes and narrower quality outcomes. Applications, job starts, quits, layoffs, and raises are asked of larger groups and help show whether the tightness measure behaves as a labor-market measure should. The amenity comparisons are more detailed but apply to a selected group. Neither group is a complete census of opportunity. Together they link the environment around workers to the transitions that can change a job’s quality.
What the measurement system leaves out
The local measure is closer to labor demand than the unemployment rate because it begins with employers’ reported recruitment. But it is not a pure measure of opportunity for every worker. State geography may be too broad for a person whose search is limited to one city or commuting corridor. Conversely, a state may contain several labor markets with very different industry mixes. A state-level average can identify the surrounding environment while hiding meaningful variation inside it.
The paper addresses some of this concern with a second source, Lightcast. Lightcast compiles online job postings on a rolling basis and offers more geographic detail, including metropolitan and micropolitan areas. It is useful as an independent check because it does not come from the JOLTS establishment survey. But the two measures differ in ways that matter. Lightcast is limited to jobs advertised online, while JOLTS includes active recruiting through other channels. Lightcast captures postings throughout the month, while JOLTS measures openings on the last business day. Lightcast does not impose the same 30-day start condition, and one posting does not always map cleanly to one opening.
Agreement between the measures therefore means something limited but valuable. If both show that tighter local conditions are associated with better reported job outcomes, the relationship is less likely to be an artifact of one data source’s exact construction. It does not mean that the two sources measure the same number, or that either one tells us how many offers a particular worker received.
The choice of geography creates another trade-off. States provide public-use identifiers and a broad labor-market environment, but they may be less precise than the area in which a worker actually searches. Core-Based Statistical Areas can better approximate local labor markets for some workers, and the paper uses Lightcast to examine that level. The estimates there are somewhat smaller, although generally not statistically different from the state-level results. Migration also complicates interpretation: if people move toward stronger markets, a local estimate can be affected by who is present in each place.
The study’s identification strategy adds another layer. The baseline approach uses variation over time and across geographies, controls for individual characteristics, and includes geography fixed effects. That design asks whether outcomes are better in places and years with more openings, after accounting for persistent differences across places. It does not make the raw opening measure a laboratory treatment. Firm decisions to post vacancies can respond to labor supply, population changes, skill mismatch, or other factors that also affect worker outcomes.
The authors therefore use a shift-share instrument as a supplementary approach. It projects national, industry-level employment changes into local areas using preexisting industry employment shares. The intention is to isolate labor-demand variation that is less likely to be caused by a local workforce’s own response. This approach produces similar, generally larger effects, but it depends on assumptions about the exogeneity of the national industry shocks. A stronger estimate under one design is not a reason to discard the assumptions that made the design informative.
Time variation is especially important. Much of the opening movement in the study occurred during the exceptional pandemic recovery. When the authors add year fixed effects, many estimates become smaller and sometimes statistically indistinguishable from zero because the controls remove the broad national movement that identifies much of the baseline relationship. This does not erase the paper’s result, but it tells us what kind of evidence it is: a study of local and temporal variation during a large labor-market transition, not a timeless constant that applies identically to every future slowdown.
There is also a difference between measuring labor demand and measuring worker welfare. A rise in openings can increase a worker’s options, but the options may be badly matched, inaccessible, or short-lived. A fall in openings can reduce bargaining power, but a worker with a strong network, portable skills, or an unusually good current employer may be little affected. Aggregate measures describe the opportunity set around people; they do not determine each person’s decision.
This is why the measurement system should be read as a chain with interfaces rather than as one master statistic. JOLTS supplies an establishment-based measure of active demand. Population data normalize it. Geography and time align it with households. SHED supplies reported transitions and comparisons of job quality. Lightcast and the shift-share design test whether the relationship survives a different measure or identification strategy. Each interface adds information, but each also adds assumptions.
The distinction becomes important when interpreting the Federal Reserve estimates. A reported association between openings per capita and better job quality is not a claim that every vacancy improves a worker’s career. It is evidence that, on average, local labor-market conditions are related to the likelihood of moving and reporting improvement. The next question is how large those relationships are, and whether the movement shows up first in search, in job changes, or in the qualities of the destination job.