Grow Now, Pay Later: What 123 "Paid to Expand" Earnings Calls Sound Like
This study profiles earnings calls whose language answered YES to the hypothesis "Paid to expand," drawing on 498 calls from 2015 to 2024, of which 123 (about a quarter) qualified. These calls ran slightly more candid (7.02 vs 6.91), more specific (7.79 vs 7.64), and more confident (7.53 vs 7.33) than the base, while showing less evasion (2.52 vs 2.71) and stress (2.19 vs 2.39). Guidance patterns were broadly similar, with 26.0% raising guidance vs 21.7% in the base. Underperformance flags appeared on calls scored as rehearsed (0.65) and CFO-dominated (0.62). Among 55 matched calls, the median forward return was -0.153 vs -0.105 in the base, with 38.2% beating expectations in both groups.
- Of 498 calls in the 2015-2024 corpus, 123 (24.7%, CI 21.1%-28.7%) answered YES to "Paid to expand."
- These calls showed higher specificity (7.79 vs 7.64), higher confidence (7.53 vs 7.33), and lower evasion (2.52 vs 2.71) than the base.
- 26.0% of paid-to-expand calls raised guidance versus 21.7% of base calls, while withdrawal rates were similar at 1.6% vs 1.0%.
- Among 55 matched calls, the median forward return was -0.153 (mean -0.103) versus -0.105 in the base, and the beat rate was essentially identical at 38.2% vs 38.2%.
1Introduction
Expansion usually shows up in capital-expenditure lines and headcount plans, but it also leaves fingerprints in how management talks. Calls where our classifier detected a "Paid to expand" posture are an opportunity to ask whether the tone of an expansion matches its economics: are these more confident, more specific conversations, and do they behave like the average call afterward? For anyone who reads earnings transcripts for a living, tone shifts during growth phases are a recurring but poorly documented pattern. This study describes 123 such calls drawn from 498 calls spanning 2015 through 2024, examining their language profile, guidance behavior, and subsequent return distribution.
2Data & methodology
The corpus comprises 498 earnings-call transcripts published between 2015 and 2024, 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 that answered YES to the research hypothesis "Paid to expand" (n = 123; 24.7% of the reference set, 95% Wilson interval 21.1%–28.7%). Baseline figures use the set of calls on which this question was tested. 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 language deltas are consistent and modest: paid-to-expand calls are more candid (7.02 vs 6.91), more specific (7.79 vs 7.64), and more confident (7.53 vs 7.33), with lower evasion (2.52 vs 2.71) and stress (2.19 vs 2.39). Two underperformance flags stand out: calls scoring 0.65 on the "Reads Rehearsed" signal and 0.62 on "CFO Dominates." The share of these calls rises from 0.05 in 2016 to 0.16 in 2022 before easing to 0.07 in 2024, suggesting expansion talk clusters in certain market regimes. Returns offer no comfort: the median of -0.153 trails the base median of -0.105, though the 38.2% beat rate matches the base exactly.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 7.02 | 6.91 | +0.11 |
| Evasion | 2.52 | 2.71 | -0.19 |
| Specificity | 7.79 | 7.64 | +0.14 |
| Stress | 2.19 | 2.39 | -0.20 |
| Promotion | 5.15 | 5.11 | +0.04 |
| Confidence | 7.53 | 7.33 | +0.20 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 26.0% | 21.7% |
| Maintained | 53.7% | 53.2% |
| Lowered | 12.2% | 13.3% |
| Withdrawn | 1.6% | 1.0% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| When the CFO Dominates | 0.62× | 8.9% | 14.5% |
| Calls That Read Rehearsed | 0.65× | 26.2% | 40.4% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -15.3% | -10.5% |
| Interquartile range | -28.2% to +6.6% | — |
| Share beating SPY | 38.2% (95% CI 27%–51%) | 38.2% |
| Observations | 55 | 199 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| MCD | Q2 2024 | 2024-07-29 | D |
| ASO | Q1 2024 | 2024-06-11 | C+ |
| NOAH | Q1 2024 | 2024-05-30 | D |
| WRBY | Q1 2024 | 2024-05-09 | A |
| AZEK | Q2 2024 | 2024-05-08 | B+ |
| LINC | Q1 2024 | 2024-05-06 | B+ |
| PPC | Q1 2024 | 2024-05-03 | A |
| TTI | Q1 2024 | 2024-05-01 | A |
4Discussion
A careful reader should conclude that calls flagged as "Paid to expand" sound measurably more confident and specific, and that their subsequent return distribution in this sample skews slightly worse than typical calls. They should not conclude that expansion language causes underperformance, that these calls predict returns, or that the pattern would persist out of sample. The two flagged signals (0.65 and 0.62) are descriptive scores from the same corpus, not validated trading indicators. The beat-rate equivalence at 38.2% versus 38.2% is the safest summary: on expectations, these calls look ordinary.
5Limitations
The language fields are AI-read and noisy, so deltas of 0.11 to 0.2 should not be over-interpreted. The returns sample covers 55 flagged calls drawn from a base of 22,449 calls skewed toward liquid names, which limits generalization to smaller issuers. Our own forward tests falsified directional prediction from these signals, and LLM scoring can partially remember famous stocks' histories, contaminating any backtest. The trend shares are computed against varying annual call counts, so year-to-year comparisons are descriptive only. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.
Companion page: every company matching this hypothesis is listed at the question’s own page.