Proprietary Intelligence Indices
Four composite indicators computed from 150+ live data sources. Each synthesizes multiple independent signal channels into a single actionable reading. Together, they form the foundation of Convex’s macro intelligence framework.
CVRP — Convex Recession Probability Index
Single 0–100 composite leading indicator
The CVRP synthesizes five independent recession signal channels into a single reading. Each channel is scored 0–20 based on how strongly it signals recessionary conditions. The composite is normalised to account for missing components — if only 4 of 5 channels have data, the index scales accordingly and flags reduced confidence.
The design philosophy is multi-channel triangulation: no single data point can push the index to extreme levels. A reading above 60 requires at least 3 of 5 channels to be signalling simultaneously.
| Channel | Source | Weight |
|---|---|---|
| Yield Curve Inversion Depth | T10Y2Y, T10Y3M | 0–20 points |
| Sahm Rule Proximity | SAHMREALTIME | 0–20 points |
| Initial Claims Momentum | IC4WSA (4-week moving average) | 0–20 points |
| High-Yield Credit Spread Z-Score | BAMLH0A0HYM2 | 0–20 points |
| Leading Economic Index | USSLIND (discontinued Feb 2020) | 0–20 points |
Uses the worse spread between 10-year minus 2-year and 10-year minus 3-month Treasuries. When either curve inverts (spread goes negative), the score escalates proportionally. Yield curve inversion has preceded every US recession since 1970, typically by 12–18 months.
Tracks acceleration in the unemployment rate versus the 0.50 percentage point threshold established by economist Claudia Sahm. A 100% accuracy rate identifying US recessions since the 1950s. Captures labour market deterioration that is already underway, not forecast.
3-month percentage change in the 4-week moving average of initial unemployment claims. The highest-frequency labour indicator available (weekly). Leads non-farm payroll data by 2–3 months, making it an early warning for labour market deterioration.
Z-score of ICE BofA high-yield option-adjusted spread versus its own 1-year rolling history — not absolute levels. This avoids secular compression bias: a spread of 400bps means something different in 2024 than in 2008. Captures the market's real-time pricing of default risk.
Conference Board LEI aggregated 10 leading indicators. Discontinued in February 2020 — the index now operates on 4 components. When fewer than 3 components are available, the index flags reduced confidence.
CVRP = (sum of component scores) / (available_components × 20) × 100
Requires at least 3 of 5 components to compute. When fewer are available, the index is marked as low-confidence. This prevents a single extreme reading from dominating.
CNLI — Convex Net Liquidity Index
Actual market liquidity after accounting for critical drains
The Federal Reserve’s balance sheet is often cited as a proxy for market liquidity, but this is misleading. Two large drains — the Reverse Repo Facility (RRP) and the Treasury General Account (TGA) — can absorb hundreds of billions of dollars that never reach financial markets. The CNLI strips these out to reveal actual market liquidity.
This distinction matters enormously. During 2022–2023 QT, the headline balance sheet was shrinking, yet CNLI actually increased because the RRP was draining faster than the Fed was tightening. This explains the paradoxical 2023 equity rally during an ostensible tightening cycle.
CNLI = WALCL − RRPONTSYD − WTREGEN
WALCL — Federal Reserve Total Assets (weekly, balance sheet)
RRPONTSYD — Overnight Reverse Repurchase Agreements (daily). Cash parked by money market funds earning risk-free rate, removed from the financial system.
WTREGEN — Treasury General Account (weekly). When Treasury builds its cash balance, it drains liquidity from the banking system.
CRAI — Convex Risk Appetite Index
Cross-asset real-time risk appetite from 5 independent price ratios
The VIX measures equity option demand — one dimension of risk sentiment. The CRAI spans five independent cross-asset price ratios, each capturing a different facet of risk appetite: equity size preference, credit quality preference, consumer spending behaviour, global risk distribution, and banking sector health.
Each ratio is converted to a 60-day rolling z-score, then mapped to a 0–20 point scale. A z-score of −2 maps to 0 points (extreme aversion), +2 maps to 20 points (extreme appetite). The composite sum is normalised to 0–100.
| Signal | Ratio | What It Captures |
|---|---|---|
| Small Cap / Large Cap | IWM / SPY | Small-cap equities carry higher economic sensitivity, less analyst coverage, and weaker balance sheets. When IWM outperforms SPY, markets are expressing confidence in growth continuation and risk-taking. |
| High Yield / Investment Grade | HYG / LQD | Direct credit market risk appetite signal. Investors actively choosing risky corporate debt over investment-grade bonds indicates willingness to accept default risk for higher yield. |
| Consumer Discretionary / Staples | XLY / XLP | Consumer spending patterns: discretionary outperformance signals spending confidence and economic optimism. Staples outperformance signals defensive positioning. |
| Emerging Markets / Developed | EEM / EFA | Global risk appetite. Emerging markets carry currency risk, political risk, and higher economic beta. EM outperformance signals that investors are reaching for growth globally. |
| Regional Banks / Broad Market | KRE / SPY | Regional bank equities are sensitive to the yield curve shape, commercial real estate exposure, and the lending cycle. KRE outperformance signals confidence in credit creation and domestic economic activity. |
points = clamp((z_score + 2) / 4 × 20, 0, 20)
Each ratio is independently z-scored against its own 60-day rolling mean and standard deviation. This means the CRAI adapts to the current volatility regime — a move that would be extreme in a calm period registers as normal during a crisis.
NVI — Narrative Velocity Index
Real-time information environment acceleration from 46 editorial sources
The NVI measures how fast the macro narrative is changing — not what the narrative is, but how quickly it is accelerating. It leads sentiment surveys and positioning data because narrative shifts precede opinion shifts, which precede position changes.
The index draws from 46 editorially diverse RSS sources spanning 7 categories: establishment (Bloomberg, FT, Reuters), contrarian outlets, wire services (AP, AFP), government communications (Fed, Treasury, ECB, BoE), academic/think tank publications, crypto-native media, and neutral sources. This diversity is the key — consensus across normally-opposed sources is a structural signal, not noise.
The NVI uses zero additional AI API calls. Term extraction uses curated regex dictionaries. Sentiment piggybacks on existing Claude evaluation calls. This makes the index extremely cost-efficient to compute every 30 minutes.
Term Frequency Acceleration
Tracks ~80 curated financial terms across 46 sources using a 7-day rolling window. Measures which narratives are gaining traction — for example, a spike in "rate cut" mentions often precedes actual policy shifts by weeks. Acceleration (rate of change of frequency) matters more than absolute mentions.
Sentiment Divergence
Measures convergence or divergence of bullish/bearish sentiment across 7 editorial categories. When normally-opposed sources (contrarian vs. establishment) converge on the same outlook, it signals genuine structural consensus. Divergence signals high uncertainty and whipsaw risk.
Source Primacy Shift
Tracks the rate of change in which sources are breaking stories first. When wire services (AP, Reuters, AFP) dominate, it indicates a fast-moving news environment — "sprint mode" for macro surprises. Distributed news indicates a quieter period.
NVI = (Term Acceleration × 0.40) + (Sentiment Divergence × 0.35) + (Source Primacy × 0.25)
Updates every 30 minutes. Requires a 14-day warm-up period for reliable readings. During the warm-up, values are flagged as preliminary.
Cross-Index Playbook
How the four indices interact to classify market regimes
Individual index readings are informative, but the real value emerges from their interaction. The Cross-Index Playbook combines all four live readings into a regime classification with specific trading implications. This runs automatically on every indicator page.
Risk-Off Confirmed
CVRP ≥ 60 + CRAI < 40 + NVI ≥ 60 + Converged Bearish
Highest conviction risk-off environment. Recession signals, defensive positioning, narrative acceleration, and editorial consensus all aligned. Reduce equities, favour Treasuries and cash.
Complacency
CVRP < 30 + CRAI ≥ 70 + NVI < 40
Low risk, high appetite, quiet narrative. Builds vulnerability to negative surprises. Historically the setup that precedes sharp corrections — positions are one-sided with no hedging demand.
Divergent: Risk Ignored
CVRP ≥ 50 + CRAI ≥ 60
Recession signals flashing but investors remain aggressive. Most dangerous configuration: the gap between fundamentals and positioning must eventually resolve, often through forced selling.
Risk-On Expansion
CVRP < 35 + CRAI ≥ 50 + CNLI > $4.5T
Fundamentals, positioning, and liquidity all aligned bullish. Stay invested, focus on bottom-up selection. Watch for early deterioration in any signal.
Recession Risk With Liquidity Support
CVRP ≥ 50 + CNLI > $5.5T
Fundamentals deteriorating but Fed still pumping liquidity. The classic "climbing the wall of worry" setup. Lean toward quality over junk within risk exposure.
Liquidity Headwind
CNLI < $4.0T
Quantitative tightening or RRP absorption creating structural headwind. Punishes liquidity-sensitive assets: growth equities, crypto, small caps. Duration becomes a drag.