Project 12: Market Volatility and GARCH-X Modeling
Book: The Big Astrology Book of Research by Renay Oshop
Source: bigastrologybook.com
🌟 Overview — What We Asked
Can planetary aspects, included as exogenous variables in GARCH models, improve volatility forecasting beyond what a standard GARCH model achieves on its own? This is the most rigorous financial astrology test in the book — and it uses the industry's own standard tools.
💡 Why This Matters
Project 02 found a statistically real but tiny correlation between planetary cycles and market volatility (r = 0.0226). But correlation is not prediction. The question that matters for anyone who might actually use this information is: does knowing the planetary positions improve my next forecast?
GARCH models are the financial industry standard for volatility prediction. They explicitly account for volatility clustering — the well-documented phenomenon where turbulent periods follow turbulent periods, and calm follows calm. GARCH-X extends this by adding external variables, directly testing whether planetary data improves predictions.
If planets contribute useful information, GARCH-X should outperform baseline GARCH on out-of-sample data. If not, we'll know precisely: adding planets makes forecasts worse.
📊 The Data
| Source | Description |
|---|---|
| Yahoo Finance (VIX) | CBOE Volatility Index, 1990–2024 |
| Yahoo Finance (S&P 500) | Daily returns |
| Swiss Ephemeris | Outer planet aspect angles — continuous cosine similarity values |
| Sample | 8,500+ trading days |
Aspects analyzed: Jupiter-Saturn, Saturn-Uranus, Saturn-Neptune, Saturn-Pluto synodic cycles.
📈 Results
Phase 1: GARCH-X — Does Adding Planets Help?
| Model | In-Sample AIC | Out-of-Sample MSE (2016–2024) |
|---|---|---|
| Baseline GARCH(1,1) | −120,451.54 | 0.00013050 |
| GARCH-X + Planetary Data | −120,449.56 | 0.00013054 |
Both metrics favor the baseline. AIC increased (worsened) by 2.0 when planetary data was added. Out-of-sample MSE also worsened slightly. By every measure, planetary variables add noise, not signal, to GARCH forecasting.
Phase 2: VIX Correlations with Planetary Aspects
Despite failing the predictive test, correlations between aspects and elevated VIX do exist:
| Aspect Type | Correlation (r) | p-value |
|---|---|---|
| Hard aspects overall | 0.295 | < 0.0001 |
| Soft aspects overall | 0.089 | 0.042 |
| Saturn-Pluto | 0.312 | < 0.001 |
Mean VIX by configuration:
| Configuration | Mean VIX | N Days |
|---|---|---|
| Grand Cross | 24.8 | 145 |
| T-Square | 21.2 | 892 |
| Grand Trine | 17.8 | 234 |
| No major pattern | 19.4 | 7,229 |
Hard aspects correlate with elevated VIX; soft aspects with lower VIX. The Saturn-Pluto correlation (r = 0.312) is the strongest single-pair result.
Notable historical coincidences:
| Event | Date | VIX | Active Aspects |
|---|---|---|---|
| COVID Crash | Mar 2020 | 82.69 | Saturn-Pluto conjunction |
| 2008 Financial Crisis | Oct 2008 | 80.86 | Saturn-Uranus opposition |
| 2011 Debt Crisis | Aug 2011 | 48.0 | Uranus-Pluto square building |
| Brexit | Jun 2016 | 25.76 | Saturn square Neptune |
🔍 The Correlation-Prediction Paradox
How can planetary aspects correlate with VIX (r = 0.295, p < 0.0001) yet worsen GARCH predictions? This apparent contradiction has a clear resolution.
Hard Saturn aspects occur during roughly 20–30% of all trading days. During those same periods, economic cycles, political events, and other catalysts also increase. The planetary aspects do not cause the volatility — they are correlated with it because both respond to longer structural cycles (economic supercycles, geopolitical cycles) that coincidentally align with 15–30 year planetary periods.
In-sample correlation exploits this historical coincidence. Out-of-sample prediction fails because:
1. Planetary variables add parameters without proportional information gain
2. Future planetary cycles don't predict which specific days will be volatile
3. The economic cycles underlying the correlation are not locked to planetary periods
The regression analysis confirms the modest scale: planetary aspects alone explain only 8.7% of VIX variance (R² = 0.087). Adding economic controls raises R² to 0.342 — showing that economic variables explain vastly more.
⚠️ Statistical Caveats
- r = 0.295 is weak: ~8.7% of VIX variance explained. Day-of-week explains more.
- Multiple testing: Dozens of aspect combinations tested. Bonferroni correction would substantially raise thresholds.
- Temporal autocorrelation: Both planetary aspects and VIX exhibit strong autocorrelation. Standard p-values likely overstate significance; Newey-West standard errors should be applied.
- Look-ahead bias: Identifying which configurations coincide with famous crises is data dredging. The test that matters is whether this information was usable before the crisis.
🌟 Conclusion
This project provides the book's most methodologically rigorous financial astrology test. Using GARCH models — the industry standard for volatility forecasting — planetary aspects make predictions worse, not better.
The statistical correlation between hard aspects and elevated VIX (r ≈ 0.295–0.312) is real but practically useless. It reflects a loose historical coincidence of economic and planetary cycles — not a mechanistic relationship. Adding planetary data increases noise and worsens out-of-sample performance.
The correlation findings, taken naively, might encourage a financial astrologer to conclude they've found something real. The GARCH out-of-sample test definitively shows they have not. This is what proper financial modeling looks like: not "does this correlate historically?" but "does knowing this improve my next prediction?"
For planetary aspects and market volatility, the answer is no.
This project does not constitute financial advice.