$ prism-api --info

Behavioral Signal
Infrastructure

Quarterly behavioral signals covering the US Russell 3000, derived from 57 psychometric dimensions scored on SEC filings and earnings transcripts. Three signals. One raw panel. Delivered via CSV, API, or S3.

7,300+

Companies

310K+

Documents

Dimensions

25 yrs

History

$ prism-api --list-products

Signal Suite

Prism operationalizes "tone from the top" into a quantitative, cross-sectional measure. All inputs are public documents. No web scraping, no PII, no MNPI.

Conduct Risk Index R3K

+21.72% alpha · Sharpe 1.50

US long/short alpha signal. Walk-forward Carhart validated on Russell 3000. t=5.66, 11/12 formation years positive, defensive beta.

Enforcement Risk Index R3K

AUC 0.839 · 13× separation

US regulatory enforcement probability. Predicts SEC and DOJ actions across 8,342 firms. Compliance screening and D&O insurance pricing.

Prosecution Risk Score S&P 500

AUC 0.944 · nested LOO CV

DOJ criminal prosecution probability. Temporal behavioral drift signal — firms losing prosocial tone in public communications.

Global Coverage SOON

International expansion

International equity markets under validation. Publication follows the same walk-forward diligence standard as the US indices.

Tier 2 — Raw Behavioral Panel

Full 57-dimension quarterly panel. 7,300+ Russell 3000 companies, 2001–2026. Build your own signals.

$ prism-api --query conduct-risk

Conduct Risk Index

Conduct risk is a regulator-defined category with $41B+ in US enforcement fines since 2019. The Conduct Risk Index quantifies exposure from the language patterns in SEC filings and earnings transcripts — before enforcement begins.

COVERAGE ....... Russell 3000+Full US equity universe excluding micro-caps. 9,155 tickers

HISTORY ........ 2013–2025Walk-forward formation years. 132-month out-of-sample window. 132 months, quarterly

CARHART_ALPHA .. +21.72%Net of Carhart 4 factors. Walk-forward, not in-sample. annualized, walk-forward

SHARPE .......... 1.50Top decile among quantitative signals globally. 132 months

T_STAT .......... +5.66p < 0.001. Survives every standard correction. p < 0.001

MAX_DRAWDOWN ... -9.37%Worst peak-to-trough. Shallow drawdown relative to alpha magnitude.

MKT_BETA ....... -0.333Negative beta = the signal is defensive. Gains when markets fall. defensive

YEARS_POS ...... 11/1211 of 12 formation years positive. Worst year: -7.58% (2014). 92%

SMB_BETA ....... -0.130Not a small-cap effect. Signal works across market cap spectrum. not size-driven

UMD_BETA ....... -0.062Not a momentum effect. Statistically insignificant. not momentum

Walk-forward Carhart 4-factor regression. Russell 3000 universe. Formation years 2013–2023.

Conduct Risk — Annual Returns

+25%

+3%

+13%

+4%

+14%

+37%

+23%

+9%

+40%

+26%

+6%

+9%

+11%

+48%

+16%

+10%

+27%

+19%

+22%

'06

'07

'08

'09

'10

'11

'12

'13

'14

'15

'16

'17

'18

'19

'20

'21

'22

'23

'24

'25

-10%

'06+25%

'07+3%

'08+13%

'09+4%

'10+14%

'11+37%

'12-10%

'13+23%

'14+9%

'15+40%

'16+26%

'17+6%

'18+9%

'19+11%

'20+48%

'21+16%

'22+10%

'23+27%

'24+19%

'25+22%

Full 20-year backtest range shown. Validated walk-forward window (formation years 2013–2023) delivers +21.72% net of Carhart 4 factors, Sharpe 1.50, 11 of 12 years positive.

$ prism-api --query enforcement-risk

Enforcement & Prosecution Risk

Firms whose public communication loses its prosocial and moral tone over time are the enforcement-prone ones. The signal is drift, not level.

PROSECUTION_AUC . 0.94Nested cross-validation on S&P 500. 495 firms, 24 prosecuted. Honest out-of-fold scores. nested CV, S&P 500

LEAD_TIME ...... ~2 quartersSignal detectable before DOJ enforcement action. before enforcement

SIGNAL_TYPE .... Behavioral driftFirms losing prosocial tone over time, not cross-sectional level. temporal, not cross-sectional

ORTHOGONALITY .. 94.7%Signal variance NOT explained by sector + log-assets. Strongly idiosyncratic. unexplained by sector + size

FEATURES ....... 15Comparable AUC with 29 total features vs 242 in full model (1/8 the feature count). parsimonious composite

BOOTSTRAP_CI .. [0.91, 0.97] 95% CI, N=1000

$ prism-api --methodology

How It Works

SIGNAL Behavioral scores from regulatory filings and earnings transcripts

METHOD Proprietary scoring methodology

UNIVERSE US Russell 3000 (7,300+ companies)

REBALANCE Annual

DIMENSIONS 57 behavioral constructs per company

SOURCES SEC EDGAR (10-K, 10-Q, DEF 14A) + earnings call transcripts

BUILT_BY PhD organizational psychologist (The New School for Social Research · Columbia Business School)

$ prism-api --pricing

Access

Evaluation

Sample data, dictionary, catalogue, and case studies.

Tier 1 — Alpha Signals

Conduct Risk Index, Enforcement Risk Index, Prosecution Risk Score. Quarterly, via API, CSV, or S3.

Tier 2 — Raw Panel