A descriptive look at how BLS JOLTS labor tightness co-moves with sector-median operating margins from 2022 to 2025. Patterns differ sharply across the seven sectors in our equity panel.
May 13, 2026
Every cycle, investors argue about whether tight labor markets are about to compress corporate margins. The structural argument is familiar: wages are sticky on the way up and labor is the largest single line item for most service businesses. When BLS reports quits rates near multi-decade highs, the consensus expectation is that operating margins should follow lower with a lag.
The argument is also structurally incomplete. Sectors differ by an order of magnitude in their labor cost share. A regional bank does not have the same labor exposure as a hospital system. A pipeline company does not look like a staffing firm. A software platform with a 65% gross margin and a fixed engineering base does not look like a contract manufacturer. Any average relationship between labor tightness and margins across all sectors will be diluted by aggregation.
This piece does not try to establish causality. We describe how a sector-level BLS JOLTS tightness index co-moves with sector-median operating margins from 2022 through 2025, sector by sector. The descriptive question is interesting on its own: where does the textbook negative relationship show up in the data, and where does it not?
Labor tightness from BLS JOLTS. The Job Openings and Labor Turnover Survey publishes monthly job openings, hires, quits, and layoffs rates for ten supersectors. We use annual averages from 2019 to 2025 for the two rates most diagnostic of tightness from the employer side. The first is the quits rate, because workers only leave when they see better options. The second is the openings rate, which measures unfilled demand. We z-score each rate within sector across the available history so the index measures deviation from the sector’s own norm, not the raw level. The composite tightness index is the equal-weighted mean of the two z-scores.
We map each JOLTS supersector to the seven sectors in our existing equity panel:
| Panel sector | JOLTS supersector |
|---|---|
| Consumer | Trade, transportation, and utilities |
| Energy | Mining and logging |
| Financials | Financial activities |
| Healthcare | Education and health services |
| Industrials | Manufacturing |
| Real Estate | Financial activities (closest proxy) |
| Technology | Information |
Mappings reflect the closest BLS supersector for each panel sector. Real Estate has no dedicated JOLTS series and is mapped to Financial activities as a proxy; results for Real Estate should be read with that caveat.
Margins from the existing equity panel. Operating margins come from the same 140-company panel used in The Pricing Power Map and the Margin Expansion Playbook. The panel combines yfinance fundamentals with AI adoption scores derived from SEC 10-K filings. We compute the sector-median operating margin per fiscal year and the year-over-year change in basis points.
The sample is short. Operating margins are populated for fiscal years 2022-2025 (four observations per sector, three lagged observations). This is enough to compare sectors against each other within a common window, but it is not enough to make a clean cyclical claim in either direction. The pooled coefficients should be read as descriptive of this window, not as out-of-sample alpha. We are honest about that throughout.
Every sector shows the same shape: a trough in 2020, a sharp spike in 2021-2022, and a steady normalization through 2024-2025. The amplitude differs. Information and Financial activities have the most pronounced spikes. Those are knowledge-worker labor markets where wage and openings dynamics moved hardest. Mining and logging is the most muted: a sector where headcount is bound to capex cycles, not labor market sentiment.
The within-sector z-score is the right normalization for this exercise. Comparing Leisure & Hospitality’s raw 5% quits rate to Information’s 2% rate and calling the former “tight” mixes structural industry differences with cyclical movement. The z-score asks instead whether each sector is currently above or below its own norm. (We return to a real limitation of this standardization, the dominance of the COVID period in the reference window, in Limitations.)
| Lag (years) | N | Correlation | Slope (bps per 1z tightness) |
|---|---|---|---|
| 0 | 21 | -0.02 | -10 bps |
| 1 | 22 | +0.19 | +52 bps |
| 2 | 15 | +0.56 | +164 bps |
Pooled across sectors with sector fixed effects (within-sector demeaning). Slope is operating margin change in basis points associated with a one-standard-deviation increase in the labor tightness index, at the specified lag.
The pooled regression with sector fixed effects shows a small negative contemporaneous correlation (essentially zero) and positive correlations at one- and two-year lags. Read literally, in this sample, sectors where labor tightened tended to see operating margins rise in the following one to two years.
That is the opposite of the textbook prediction. It is also unsurprising given the window. The four years from 2022 through 2025 are dominated by the post-COVID recovery: labor markets tightened during the same period that margins normalized off 2020 lows. The two variables co-moved up and then co-moved down. A single pooled coefficient over this window mostly measures that shared cycle.
That is why we lean on the sector-by-sector decomposition rather than the pooled number for the rest of the piece, while noting that the same small-sample caveat applies to every sector-level correlation.
Healthcare sits alone on the negative side of the chart, with a correlation near -0.97. The Education and Health Services supersector is one of the most labor-intensive parts of the US economy, with wages and salaries running above 50% of operating costs at most hospital systems. The descriptive pattern in our sample is consistent with that structural story: years when the JOLTS quits and openings rates ran hot in this supersector are also years when sector-median operating margins compressed. The directional fit is striking on its own terms. It also rests on three observations, and we are careful in the limitations section about how far that can be pushed.
The other six sectors all show positive correlations. We are not claiming labor does not affect their margins. We are observing that, in this window, other factors moved with the labor cycle and dominated the descriptive picture:
The simple read is that, outside Healthcare, this window does not isolate a labor signal from the broader macro mosaic. With more years and a richer set of controls, we may find that labor exposure shows up cleanly in other sectors too. With this sample, we cannot say.
Healthcare’s negative correlation is strongest contemporaneously (lag 0) and persists at lag 1, then loses sample at lag 2 (n drops below 3). The contemporaneous shape is consistent with how healthcare wage pressure tends to be discussed in the literature: contract renegotiations, agency-nurse premia, and benefits inflation typically hit reported operating expenses within the same fiscal year. We are not testing that mechanism here; we are noting that the descriptive shape is consistent with it.
In Energy, Financials, Real Estate, and Technology, the positive correlation strengthens at lag 1. That pattern is more consistent with shared cyclical co-movement than with a labor-cost mechanism, since there is no plausible story under which a tighter labor market causes margins to rise. We read it as descriptive evidence that JOLTS is acting as a coincident cycle indicator in those sectors over this window.
These are observations from the descriptive analysis, not investment recommendations. We are deliberately not translating three-observation correlations into sector tilts.
Healthcare is the only sector where the negative co-movement appears in this window. That is the sector where the textbook labor-cost story is strongest on a priori grounds, and the descriptive picture is consistent with it. It is also where the sample is thinnest. If labor tightness re-accelerates and Healthcare margins compress, this sample will look like a leading indicator; if they do not, the relationship was likely confounded.
In the other six sectors, JOLTS in this window behaves more like a coincident cycle indicator than a labor signal. Reading the BLS data as a labor-specific margin driver for Energy, Financials, Real Estate, or Technology over 2022 to 2025 is over-interpretation given the shared cyclical drivers.
The within-sector view matters more than the pooled view. A single coefficient across sectors averages away the cost-structure heterogeneity that the descriptive picture suggests is doing most of the work.
The binding constraint is sample length. As history extends beyond a single post-COVID cycle, the descriptive picture may look very different in every sector, including Healthcare.
This piece is descriptive on a short, atypical sample. The list below names the things we do not claim, and where each one points for future work.
Descriptive, not causal. We report correlations between labor tightness and operating margins. We do not identify a causal labor-cost channel. With shared cyclical drivers in our window (post-COVID demand recovery, supply chain normalization, Fed tightening), the same data is consistent with multiple structural stories. Anything stronger would require an identification strategy (instrument, event study, controlled comparison) that we do not provide.
Sample length and statistical inference. Four annual observations per sector, three with valid lag-1 data, two with lag-2. Pooled n caps at 22. With samples this small, formal inference statistics would be misleading. We deliberately do not report standard errors, confidence intervals, or p-values, because none of them would be informative at this n. Read every correlation in this piece as a directional pattern in a small sample, not as a population estimate.
Multiple testing. Seven sectors crossed with three lags yield 21 correlations. At least one strong reading is expected by chance even under the null of no relationship. Healthcare’s -0.97 is consistent with the textbook story; it is also exactly the kind of outlier this design tends to produce, and we have not corrected for multiple comparisons.
Index construction. The tightness index z-scores each sector’s quits and openings rates over the 2019 to 2025 window. That window is dominated by the COVID trough and post-COVID spike, so the “deviation from sector norm” the index measures is, in practice, distance from the post-COVID phase. Standardizing over a longer pre-COVID baseline (JOLTS data extends back to December 2000) would treat 2021 to 2022 as a true outlier against a stationary reference distribution rather than as half of the reference distribution itself. This is a real limitation of the index, not a notation choice.
Frequency. JOLTS is monthly. We aggregate to annual averages to match the equity panel, which discards most of the within-year variation in labor markets and prevents finer tests of the lag structure. The natural next step is a panel of quarterly operating margins from 10-Q filings paired with quarterly JOLTS averages, which would multiply the effective sample size and let the analysis distinguish short lags from coincident co-movement.
Sector mapping. Real Estate is mapped to Financial Activities, which understates labor cost share for property managers and overstates it for asset-heavy REITs. Consumer maps to Trade, Transportation, and Utilities, which lumps Walmart, FedEx, and a utility. Technology maps to Information, which misses the semis (BLS Manufacturing) and IT services (Professional and Business Services) inside our tech cohort. We do not report sensitivity to alternative mappings in this version.
Cycle confound. The 2022 to 2025 window blends post-COVID recovery, supply chain normalization, and Fed tightening. Without a longer panel covering at least one independent labor cycle, labor and growth cannot be cleanly disentangled.
No company-level overlay. A natural extension is to score single names within each sector by estimated labor cost share, defined as headcount times sector average wage divided by revenue, and look at the within-sector dispersion in margin response. We will add this overlay in a follow-up.
| Sectors | 7 | panel-to-BLS mapped |
| Years | 5 | 2022-2026 |
| Sector-year observations with lag-1 valid | 22 | used in pooled regression |
| Tightness index | Equal-weighted z-score of quits rate and openings rate, standardized within sector across 2019-2025 | |
| Margin variable | Sector-median operating margin (yfinance), year-over-year change in basis points | |
| Lags tested | 0, 1, 2 years | |
Labor tightness index, for sector \(s\) and year \(t\):
\[\text{Tightness}_{s,t} = \tfrac{1}{2}\bigl(z(\text{Quits}_{s,t}) + z(\text{Openings}_{s,t})\bigr)\]
where \(z(\cdot)\) standardizes within sector across the 2019-2025 window.
Pooled regression with sector fixed effects via within-sector demeaning:
\[\Delta\text{Margin}_{s,t} - \overline{\Delta\text{Margin}}_s = \beta \cdot \bigl(\text{Tightness}_{s,t-k} - \overline{\text{Tightness}}_s\bigr) + \varepsilon_{s,t}\]
for lags \(k \in \{0, 1, 2\}\).
By-sector test, a simple OLS within each sector:
\[\Delta\text{Margin}_{s,t} = \alpha_s + \beta_s \cdot \text{Tightness}_{s,t-k} + \varepsilon_{s,t}\]