The Permanent Portfolio
WeightingHold 25% SPY (stocks), 25% TLT (long Treasuries), 25% GLD (gold) and 25% SHV (short-term Treasuries, the cash sleeve), rebalanced back to equal quarters on the first trading day of each January. Benchmark is 100% SPY.
See this in the terminal — live earnings bands, fair value and rates.
Start free trialRead the story
Harry Browne's four-way split is built never to lose badly. It also never wins big.
The rule. Hold 25% SPY (stocks), 25% TLT (long Treasuries), 25% GLD (gold) and 25% SHV (short-term Treasuries, the cash sleeve), rebalanced back to equal quarters on the first trading day of each January. Benchmark is 100% SPY.
What actually happened.
- $10,000 became $11,194. The same $10,000 in S&P 500 (SPY) became $84,528. This strategy trailed the benchmark by $73,335.
- Compounded at 0.8% a year vs 15.3% for the benchmark.
- Best year: 2014 (+9%). Worst year: 2013 (-2%). This is not a smooth ride.
The catch. At its worst, this strategy fell -8% from a prior peak — shallower than the benchmark's -34%. Would you have held through that? Most people don't. That's the whole game.
The takeaway. The lesson isn't "this idea is stupid." It's that the simple, boring benchmark is harder to beat than the pitch decks admit. Before you try to be clever, respect the index.
Drawdown from prior peak
Calendar-year returns
| Line | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Permanent Portfolio (25/25/25/25) | -0% | +6% | -2% | +9% | 0% | -0% | -0% | -0% | -0% | -0% | -0% | -0% | -0% | -0% | -0% | -0% |
| S&P 500 (SPY) | +8% | +16% | +32% | +13% | +1% | +12% | +22% | -5% | +31% | +18% | +29% | -18% | +26% | +25% | +18% | +13% |
Browne's original sleeves are stocks, long bonds, gold and cash; SHV (1-3 month Treasuries) stands in for cash, so this line does earn the T-bill yield rather than nothing. How we test
Rebalance: annual (first trading day of January), back to 25% each · Window: 2011-09-02 → 2026-09-02 · Last run Sep 2, 2026 · JSON