Golden Butterfly: the Risk X-Ray
The Golden Butterfly takes Browne's Permanent Portfolio and tilts one wing toward prosperity by adding a fifth slice of small caps. PortfolioCharts built it by optimizing for consistency of real returns across historical decades rather than for the highest peak. The analysis below shows what the five-way symmetry looks like through a risk lens, which slices actually move together, and what the small-cap tilt adds in beta.
Tyler, PortfolioCharts.com (2016)
Analysis as of Friday 2026-08-28 · recomputed weekly from daily market data · methodology
What the Golden Butterfly holds
The published allocation: 20% total US stocks, 20% small-cap value, 20% long-term Treasuries, 20% short-term Treasuries, 20% gold.
| Fund | Measured role | Weight | Expense ratio |
|---|---|---|---|
| VTI Vanguard Total Stock Market ETF | correlated core | 20% | 0.03% |
| IWM iShares Russell 2000 ETF | risk driver | 20% | 0.19% |
| TLT iShares 20+ Year Treasury Bond ETF | shock absorber | 20% | 0.15% |
| SHY iShares 1-3 Year Treasury Bond ETF | shock absorber | 20% | 0.15% |
| GLD SPDR Gold Shares | correlated core | 20% | – |
The canonical small-cap value sleeve is implemented with IWM (small-cap blend), the closest covered ETF; the original uses a dedicated small-cap value fund.
Where the risk actually sits
Measured through the covariance matrix, IWM dominates the risk budget at 35.7% of total risk for 20.0% of capital; at the other end, SHY accounts for only 1.4%.
| Fund | Share of risk | Risk | Capital | Beta | Role |
|---|---|---|---|---|---|
| IWM | 35.7% | 20.0% | 1.06 | risk driver | |
| VTI | 25.1% | 20.0% | 1.01 | correlated core | |
| GLD | 22.8% | 20.0% | 0.18 | correlated core | |
| TLT | 15.0% | 20.0% | 0.11 | shock absorber | |
| SHY | 1.4% | 20.0% | 0.0 | shock absorber |
How the pieces move together
VTI + IWM move as one block (average correlation 0.82): together they are 40.0% of the capital and 60.8% of the risk.
TLT + SHY move as one block (average correlation 0.72): together they are 40.0% of the capital and 16.4% of the risk.
| Holding | VTI | IWM | TLT | SHY | GLD |
|---|---|---|---|---|---|
| VTI 20% | 1.00 | ||||
| IWM 20% | 0.82 | 1.00 | |||
| TLT 20% | 0.13 | 0.21 | 1.00 | ||
| SHY 20% | 0.05 | 0.15 | 0.72 | 1.00 | |
| GLD 20% | 0.14 | 0.13 | 0.13 | 0.20 | 1.00 |
Pearson correlation of weekly returns · 156-week common window ending 2026-08-28 · pairbook.io
The worst it would have been
Over the analysis window, this mix would have fallen at worst -6.9% (from the week of 2026-03-06 to 2026-04-03), shallower than the S&P 500's -16.9% over the same weeks, and it has since recovered that peak.
Worst weeks of the mix: -3.3% (week ending 2025-04-11), -3.0% (week ending 2026-03-27), -2.7% (week ending 2024-11-22).
FAQ
How is this different from the Permanent Portfolio?
One slice: the 20% small-cap addition shifts the stock allocation from a quarter to 40% and roughly doubles the equity risk contribution. The comparison is visible directly in the risk-budget tables of the two pages.
Does the small-cap slice really behave differently from total-market?
Their correlation is high, as the matrix above shows, so the diversification argument rests on the size premium rather than on decorrelation. The page treats them as one block when the data says they move as one.
Run this analysis on your own portfolio
The same X-ray, risk contributions, correlation blocks, ETF overlap, runs on any mix of 4,700+ US stocks and ETFs: paste your holdings into the free portfolio X-ray, or ask an AI assistant with the PairBook MCP server.
