Free Earnings Call Evasiveness API

When management dodges analyst questions, it's often a leading indicator of trouble. Our AI scores how directly executives answer questions on a 1-10 scale.

Endpoint

GET https://securitiesdb.com/api/v1/stocks/{ticker}/earnings-transparency

How It Works

We analyze earnings call transcripts using large language models to score management responses on a 1-10 evasiveness scale:

1-3: Very Direct — management gives clear, quantitative answers
4-6: Somewhat Evasive — vague language, redirections
7-8: Evasive — obvious deflections, scripted non-answers
9-10: Highly Evasive — actively avoids answering

The API also returns the most evasive quote from the call, a trend over quarters, and an AI analysis of the deflection pattern.

Python Example — Flag Evasive Companies

import requests

watchlist = ["AAPL", "TSLA", "META", "AMZN", "GOOGL"]
flagged = []

for ticker in watchlist:
    r = requests.get(f"https://securitiesdb.com/api/v1/stocks/{ticker}/earnings-transparency")
    if r.status_code != 200:
        continue
    d = r.json()["data"]

    score = d["summary"]["latest_score"]
    trend = d["summary"]["trend"]

    if score >= 7 or trend == "becoming_more_evasive":
        flagged.append({
            "ticker": ticker,
            "score": score,
            "label": d["summary"]["latest_label"],
            "trend": trend,
            "quote": d["latest_deflection_quote"][:80] + "..."
        })

print(f"Flagged {len(flagged)} companies for evasiveness:")
for f in flagged:
    print(f"  {f['ticker']}: {f['score']}/10 ({f['label']}) — Trend: {f['trend']}")
    print(f"    Quote: \"{f['quote']}\"")

Research Background

Academic research has shown that linguistic features of earnings calls predict future stock performance. Companies with more evasive management tend to:

  • Miss future earnings estimates at higher rates
  • Have greater negative earnings surprises
  • Show declining operational performance
  • Experience increased analyst downgrades

Sources: Larcker & Zakolyukina (2012), Hobson et al. (2012), Cohen et al. (2020)