On 14.1% of calls the CFO dominates — and raised guidance runs 17.7% vs 21.1%

1Introduction

Anyone who parses earnings calls knows the CEO-CFO dynamic matters: the chief financial officer owns the numbers, and who carries the call shapes what gets said. If CFO-dominated calls are less promotional and more scripted, listeners who lean on management tone may be reading a different signal than they assume, and teams scoring call language should know which features move with the format. The pattern is common — 14.1% of all calls. This study examines those 23,336 calls within a 165,182-call corpus spanning 1990-2026, where the model answered YES to the battery item 'When the CFO Dominates,' comparing their language profile, guidance mix, phrase rates, annual prevalence, and returns sample against the full corpus.

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 the model answered YES to the battery item "When the CFO Dominates" (n = 23,336; 14.1% of the reference set, 95% Wilson interval 14.0%–14.3%). 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 behavioral profile stands out: promotion runs 4.46 vs 5.05 (-0.59) and confidence 6.95 vs 7.21 (-0.26), while specificity (7.71 vs 7.56, +0.15) and stress (2.58 vs 2.43, +0.15) both tick up — a numerate, tenser tone. The guidance mix differs too: raised 17.7% vs 21.1%, maintained 53.0% vs 48.8%, lowered 13.1% vs 11.6%. Phrase rates sharpen the picture: 'Calls That Read Rehearsed' over-indexes at 1.34x (51.8% vs 38.7%), while market-size hype runs cold — 'Scale-Dependent Advantage Claims' at 0.62x, 'A Tiny Fraction of the Market' at 0.64x, 'Volume About to Step Up' at 0.65x. Prevalence drifts from 16.42% in 2015 to 12.74% in 2025, bottoming at 11.72% in 2021. In returns, the beat flag runs 42.7% vs 39.5% and median return -4.75% vs -7.16% (n=3,556 vs 22,449).

4Discussion

The safest reading is descriptive: calls where the CFO dominates sound different — less promotional, more specific, more rehearsed — and their guidance mix leans conservative. None of this shows the CFO causes the tone, or that tone causes anything. Companies may simply hand the CFO the microphone in quieter quarters, in regulated industries, or when the story is operational rather than visionary. The higher beat flag (42.7% vs 39.5%) and less negative median return (-4.75% vs -7.16%) are single-sample comparisons, not an edge to trade on. Read this as a catalog of how call character varies with who leads it.

5Limitations

The model's YES/NO reads are noisy, and any single battery item misclassifies calls. The returns comparison rests on 22,449 calls with available data, skewed toward liquid names, so the 42.7% vs 39.5% beat rates and return medians may not generalize. Our own forward tests falsified directional prediction from these features — nothing here is an edge. Finally, LLMs partially remember famous stocks' histories, contaminating any backtest that mixes model labels with known outcomes. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.