Waiting for the Other Shoe: Calls Where It Never Drops
We study earnings calls that answered YES to the hypothesis 'The feared thing keeps not happening' — calls where management describes a widely anticipated risk that persistently fails to materialize. Of 994 calls spanning 2015–2025, 93 (9.4%) fit this pattern. Relative to the base corpus, these calls show slightly higher confidence (7.65 vs 7.36), specificity (7.78 vs 7.65), and promotion (5.20 vs 5.07), with near-identical candor and evasion. Guidance was raised on 36.6% of these calls versus 23.4% in the base. Language labeled 'Deferred Revenue Growing' appears 1.86x more often, while 'Scale-Dependent Advantage Claims' appears at 0.53x the base rate. Among the 51 calls with return data, the median 6-month return was -6.5%, essentially matching the -6.9% base median. The pattern is descriptive, not predictive.
- 93 of 994 calls (9.4%, 95% CI 7.7%–11.3%) answered YES to 'the feared thing keeps not happening'.
- Confidence scores run higher than the base corpus (7.65 vs 7.36), while candor (6.96 vs 6.93) and evasion (2.68 vs 2.67) are nearly unchanged.
- Guidance was raised on 36.6% of these calls versus 23.4% of base calls, and lowered on 8.6% versus 12.7%.
- The 'Deferred Revenue Growing' language pattern is overrepresented at 1.86x, while 'Scale-Dependent Advantage Claims' is underrepresented at 0.53x.
- The 51 calls with measured returns show a median of -6.5%, versus -6.9% for the 384-call base — a negligible difference.
1Introduction
Every earnings call has a doom that management spends the session not confirming: the feared churn wave, the looming margin squeeze, the regulatory hammer. Some companies spend years telling investors the worst-case scenario keeps not arriving. That framing is interesting precisely because it sits between reassurance and evasion — is management describing genuine resilience, or managing a narrative? For anyone who parses earnings calls for signal, knowing how often this 'the feared thing keeps not happening' posture appears, and what its language looks like, is useful context. This study measures how common the posture is, how its linguistic profile and guidance behavior compare with the broader corpus, and how its subsequent returns compare with everything else.
2Data & methodology
The corpus comprises 994 earnings-call transcripts published between 2015 and 2025, 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 "The feared thing keeps not happening" (n = 93; 9.4% of the reference set, 95% Wilson interval 7.7%–11.3%). 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 confidence delta stands out: 7.65 versus 7.36, the largest profile gap in the study, while candor and evasion are essentially flat. Guidance behavior leans optimistic — 36.6% raised versus 23.4% base, and only 8.6% lowered versus 12.7%. Language patterns point toward substance: 'Deferred Revenue Growing' appears 1.86x more often, consistent with a recurring-revenue cushion behind the un-materialized fear, while rehearsed-sounding calls (0.66x) and scale-advantage claims (0.53x) are rarer. The trend is volatile — 0.14 in 2022, 0.11 in 2023, dropping to 0.02 in 2024 and 0.0 in 2025 — suggesting the posture clusters in stressed periods. Returns show no separation: median -6.5% versus -6.9% base, with 39.2% beating versus 41.4%.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.96 | 6.93 | +0.02 |
| Evasion | 2.68 | 2.67 | +0.01 |
| Specificity | 7.78 | 7.65 | +0.13 |
| Stress | 2.40 | 2.36 | +0.04 |
| Promotion | 5.20 | 5.07 | +0.13 |
| Confidence | 7.65 | 7.36 | +0.29 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 36.6% | 23.4% |
| Maintained | 50.5% | 51.8% |
| Lowered | 8.6% | 12.7% |
| Withdrawn | 1.1% | 1.1% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Deferred Revenue Growing | 1.86× | 16.1% | 8.7% |
| The Hidden Segment | 1.29× | 28.0% | 21.7% |
| Scale-Dependent Advantage Claims | 0.53× | 4.3% | 8.0% |
| Calls That Read Rehearsed | 0.66× | 25.8% | 38.8% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -6.5% | -6.9% |
| Interquartile range | -20.6% to +12.0% | — |
| Share beating SPY | 39.2% (95% CI 27%–53%) | 41.4% |
| Observations | 51 | 384 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| AVTR | Q2 2024 | 2024-07-26 | C+ |
| IFS | Q1 2024 | 2024-05-14 | C+ |
| AZEK | Q2 2024 | 2024-05-08 | B+ |
| GL | Q1 2024 | 2024-04-23 | F |
| MNKD | Q4 2023 | 2024-02-27 | C |
| NICE | Q4 2023 | 2024-02-22 | B+ |
| TYGO | Q4 2023 | 2024-02-14 | F |
| MET | Q4 2023 | 2024-02-01 | B+ |
4Discussion
A careful reader should conclude that this posture is moderately common, slightly more confident in tone, and modestly more likely to accompany raised guidance. They should not conclude the posture is protective or dangerous: the return distributions are nearly indistinguishable (median -6.5% vs -6.9%; 39.2% vs 41.4% beats). No causal claim is available — companies that say the feared thing keeps not happening may differ in many unmeasured ways from those that don't. The 2022–2023 clustering and the 2024–2025 drop-off are observations about the sample, not forecasts. Treat every comparison here as descriptive texture about language, not a screen for outperformance.
5Limitations
All fields are AI-read labels on transcript text, so category assignment and scores carry label noise. The returns subset covers 51 of 93 calls against a 384-call base drawn from 22,449 measured calls, skewed toward liquid names, so survivorship and liquidity biases apply. Our own forward tests falsified directional prediction from these fields, and LLMs partially remember famous stocks' histories, contaminating any backtest. The 2025 bucket covers only 6,012 corpus calls and partial-year data, so the apparent trend decline may be an artifact. 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.