Trimming the Fat, Loudly: Headcount Cuts on Earnings Calls
This study examines 28,971 earnings calls where headcount was read as cutting, drawn from a 165,182-call corpus spanning 1990 to 2026. These calls make up 17.5% of the corpus (95% CI 17.4% to 17.7%). Management tone shifts sharply on cutting calls: stress rises 0.76 points and confidence falls 0.56 points versus the rest of the corpus, while specificity barely moves. Guidance behavior diverges too: 20.4% of cutting calls lowered guidance versus 11.6% elsewhere, and 6.8% withdrew it versus 2.7%. Overrepresented language themes include The Hidden Segment (1.67x) and Underused Fixed Costs (1.56x). Among 3,138 calls with return data, the median post-call return was -7.7% versus -7.2% for the base.
- Cutting-headcount calls show stress 0.76 points higher and confidence 0.56 points lower than the rest of the corpus.
- 20.4% of these calls lowered guidance and 6.8% withdrew it, versus 11.6% and 2.7% for other calls.
- The theme The Hidden Segment appears 1.67x more often on cutting calls, and Underused Fixed Costs 1.56x more often.
- Among 3,138 cutting calls with post-call returns, the median was -7.7%, slightly below the -7.2% base median.
1Introduction
When a management team starts talking about headcount, listeners lean in: staffing language often signals how leadership sees the next few quarters, and the way cuts are framed can reveal whether a company is tightening around a plan or improvising under pressure. Earnings-call followers have long treated workforce rhetoric as a tell, but the pattern is rarely measured across a large corpus. This study examines 28,971 calls, 17.5% of a 165,182-call library covering 1990 to 2026, where headcount was read as cutting, describing their tone profile, guidance behavior, recurring language themes, and post-call return distribution.
2Data & methodology
The corpus comprises 165,182 earnings-call transcripts published between 1990 and 2026, each scored independently by a large language model on an identical 37-field battery: seven categorical business verdicts, eight 0–9 behavioral meters, and twenty yes/no judgments. The study group is defined as calls where headcount was read as cutting (n = 28,971; 17.5% of the reference set, 95% Wilson interval 17.4%–17.7%). Baseline figures use all scored calls. Market outcomes join a fixed sample of 22,449 calls with twelve-month total returns in excess of SPY, measured from the first close after each call; this sample skews toward liquid U.S. names and is reported as descriptive history only.
3Results
The tone profile is the clearest signal: cutting calls run stress at 3.19 versus 2.43 elsewhere and confidence at 6.65 versus 7.21, while candor is slightly higher (6.96 versus 6.86) and specificity essentially unchanged. Guidance skews negative, with 20.4% lowered and 6.8% withdrawn against 11.6% and 2.7% in the base. Theme lifts cluster on cost logic: Underused Fixed Costs (1.56x), Scale-Dependent Advantage Claims (1.55x), and Results Worse Than Direction (1.49x). The annual trend is cyclical, peaking at 27.57% of calls in 2020, collapsing to 8.12% in 2021, and recovering to 20.54% in 2023. Returns are slightly weaker: a -7.7% median versus -7.2% base, with 40.0% beating versus 39.5%.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.96 | 6.86 | +0.10 |
| Evasion | 2.87 | 2.70 | +0.18 |
| Specificity | 7.48 | 7.56 | -0.07 |
| Stress | 3.19 | 2.43 | +0.76 |
| Promotion | 4.79 | 5.05 | -0.26 |
| Confidence | 6.65 | 7.21 | -0.56 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 11.7% | 21.1% |
| Maintained | 45.8% | 48.8% |
| Lowered | 20.4% | 11.6% |
| Withdrawn | 6.8% | 2.7% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| The Hidden Segment | 1.67× | 35.3% | 21.1% |
| Underused Fixed Costs | 1.56× | 65.2% | 41.6% |
| Scale-Dependent Advantage Claims | 1.55× | 17.1% | 11.1% |
| Results Worse Than Direction | 1.49× | 76.2% | 51.1% |
| The Finished-Story Tell | 1.45× | 6.4% | 4.4% |
| Deferred Revenue Growing | 0.66× | 5.8% | 8.9% |
| Skeptic Reassured | 0.71× | 47.0% | 66.4% |
| Volume About to Step Up | 0.73× | 20.9% | 28.5% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -7.7% | -7.2% |
| Interquartile range | -27.8% to +13.2% | — |
| Share beating SPY | 40.0% (95% CI 38%–42%) | 39.5% |
| Observations | 3,138 | 22,449 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| MOG.A | Q3 2025 | 2025-07-25 | B+ |
| ULH | Q2 2025 | 2025-07-25 | C+ |
| TNET | Q2 2025 | 2025-07-25 | C+ |
| BAH | Q1 2026 | 2025-07-25 | C+ |
| VLOUF | Q2 2025 | 2025-07-25 | C |
| PUBL | Q2 2025 | 2025-07-25 | C |
| VWAGY | Q2 2025 | 2025-07-25 | C |
| PUBL | Q2 2025 | 2025-07-25 | C+ |
4Discussion
A careful reader should treat these numbers as description, not diagnosis. Cutting calls are, on average, tenser and more guidance-negative than other calls, and their language leans on cost-and-efficiency framing. That is a correlation in the same call, not proof that the cuts caused anything or that the tone predicts returns. The return gap is small and the beat rate is nearly identical to the base. The right conclusion is that headcount-cutting calls have a recognizable texture, not that the texture is an edge.
5Limitations
Headcount-cutting is an AI-read label, and such fields are noisy; misreads cut both ways. The returns sample covers only 3,138 of these calls against a base of 22,449, skewed toward liquid names, so medians may not generalize. Our own forward tests falsified directional prediction, and LLMs partially remember famous stocks' histories, contaminating any backtest. Theme lifts come from the same noisy read pipeline. All figures describe the corpus; none establish cause, effect, or a trading edge. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.