Methodology

Overview

The Crash Gauge Dashboard monitors up to 14 economic and valuation indicators to estimate the likelihood of a significant market downturn. Each indicator is evaluated against its own 30-year historical range to determine whether current conditions are normal, elevated, or extreme.

The US dashboard uses all 14 indicators. Other countries use 6–7 indicators depending on data availability. The Developed World aggregate uses 3 global proxy indicators. All data comes from real public sources; no simulated data is used.

The 14 Indicators

Indicators span four categories:

Valuation (US only)

• CAPE (Shiller PE Ratio): Cyclically adjusted price-to-earnings ratio. High values suggest overvaluation. US only (no free non-US API).
• Buffett Indicator: Total market cap to GDP ratio. US only.

Credit & Yield

• Yield Curve Spread: Difference between long-term and short-term government bond yields. The US uses the 10Y-2Y Treasury spread; other countries use the 10Y-3M spread from OECD data on FRED. Both measures are equally valid recession signals; inversions have preceded every US recession since 1970. All 8 countries.
• Credit Spread (IG): Investment-grade corporate bond spread: US Treasury HQM Corporate Bond 10Y Par Yield minus 10Y Treasury Constant Maturity. Both series are public-domain US government data. US only.
• Central Bank Rate: Policy interest rate proxy: Fed Funds Rate (US), ECB Main Refinancing Rate (DE/FR/IT), OECD overnight interbank rate (UK/JP/CA/AU). All 8 countries.

Economic Activity

• Unemployment Rate: National unemployment rate from national statistics / OECD. All 8 countries.
• PMI (Business Confidence): OECD Business Confidence Index. Values below 100 signal contraction. All 8 countries + Dev. World.
• Consumer Confidence: OECD Consumer Confidence Index (amplitude-adjusted, 100 = long-term average) for all countries. US, UK, DE, FR, IT, JP, AU.
• Leading Economic Index: OECD Composite Leading Indicator (CLI, amplitude-adjusted) from FRED for all regions. CLI oscillates around 100 (above = expansion, below = contraction). All 8 countries + Dev. World (G7 CLI).
• Housing Starts YoY: Year-over-year change in new residential construction. US only (non-US series discontinued).
• CPI YoY: Year-over-year consumer price inflation from national CPI indices. All 8 countries.

Monetary & Market Stress

• Financial Stress Index: St. Louis Fed Financial Stress Index (STLFSI4). Constructed from 18 weekly financial data series. Zero represents normal conditions; positive values indicate above-average stress. US + Dev. World (as global proxy).
• M2 Money Supply YoY: Year-over-year growth in broad money supply. US only (non-US OECD M2 series discontinued on FRED).
• Debt to GDP: Federal government debt as a percentage of GDP. US only (non-US series only annual on FRED).

Regional Coverage

The dashboard covers 9 regions. Indicator availability varies by country:

Calculation Algorithm

The crash probability is computed in four stages:

1. Historical Percentile Ranking

Each indicator's current value is ranked against its full historical range over the selected time period. A percentile of 95 means the current value is higher than 95% of all historical observations. For indicators where higher values signal more risk (e.g., financial stress, credit spreads), a high percentile translates directly to high risk. For inverse indicators (e.g., consumer confidence, PMI), the percentile is inverted.

2. Weighted Combination

Each indicator contributes to the composite score based on its empirical relevance as a recession predictor. The yield curve, CAPE, Buffett Indicator, credit spread, and financial stress index receive the highest weights. The weighted average produces a base risk score from 0 to 100.

3. Extreme Clustering Amplification

When multiple indicators simultaneously show extreme readings (above the 90th percentile), the base score is amplified. Having 5+ indicators at extreme levels is far more concerning than just one; this non-linear clustering effect captures the compounding nature of systemic risk.

4. Sigmoid Compression

The final score is passed through a sigmoid function to compress it into a probability-like 0–100% range. This ensures the output is intuitive: values below 30% are low risk, 30–60% moderate, 60–80% elevated, and above 80% extreme.

Data Sources

All data comes from public, freely available sources:

• FRED (Federal Reserve Economic Data): primary source for most indicators across all countries
• OECD: Business Confidence Index and Consumer Confidence Index via FRED (CC BY 4.0)
• Robert Shiller, Yale University: Shiller CAPE PE ratio (US only), from the freely-published Irrational Exuberance dataset (ie_data.xls)

Data is cached server-side for up to one hour to reduce API load and improve response times. Most economic indicators are updated monthly by their respective agencies.

Limitations

• The model uses historical patterns that may not repeat in future crises.
• Non-US regions have fewer indicators (6–7 vs 14 for the US). The Developed World aggregate uses only 3 indicators and should be interpreted with additional caution.
• Economic data is typically released with a 1–3 month lag.
• The dashboard does not account for geopolitical events, policy announcements, or other qualitative factors.
• This is a statistical tool, not a crystal ball. Treat the probability as a risk barometer, not a prediction.