Swensen (Yale Endowment) vs Permanent Portfolio (Browne)
Compare two portfolios with editable holdings and named presets →These two allocations answer the same question from opposite ends. The endowment model reaches for return through illiquidity and manager access. The Permanent Portfolio gives up return in exchange for holding something that works in every economic condition. One assumes a permanent horizon and privileged access, the other assumes neither.
Set both to the same 12.8% volatility and Swensen (Yale Endowment) comes out ahead, returning 7.21% against Permanent Portfolio (Browne)'s 6.75%, a gap of 0.46 percentage points a year.
These two views disagree: the forward-looking assumptions favor Swensen (Yale Endowment), while the record since November 2006 favors Permanent Portfolio (Browne). That is worth more than either number on its own. The historical figure is one path through one set of decades, and this window contains two equity bear markets and a long gold bull run. The forward-looking figure starts from today's valuations and yields, which is the situation you are actually investing from.
Side by side
| At its own risk level | Swensen (Yale Endowment) | Permanent Portfolio (Browne) |
|---|---|---|
| Expected return | 7.21% | 5.36% |
| Volatility | 12.78% | 6.40% |
| Sharpe ratio | 0.38 | 0.38 |
Forward-looking estimates over a long horizon, not a measurement of any past period. Built from J.P. Morgan’s 2026 Long-Term Capital Market Assumptions, annualized and compounded, with a 3.10% risk-free rate. The figures are in US dollars. J.P. Morgan publishes the same assumptions in 5 base currencies and the numbers differ between them, since the expected move in the currency is part of the forecast. This page is built once, on the dollar edition; the full app runs on whichever one you choose.
The same comparison at matched risk
The table above compares two portfolios that take different amounts of risk, so part of any gap is just that. Below, the quieter portfolio is levered until both sit at 12.78% volatility. Exposure above one hundred percent is financed at cash plus 0.50 percentage points; anything under it earns the cash rate.
| Both at 12.78% volatility | Exposure | Expected return |
|---|---|---|
| Swensen (Yale Endowment) | 100% | 7.21% |
| Permanent Portfolio (Browne) | 200% | 6.75% |
Leverage is applied to the whole portfolio and assumes it can be financed at the stated rate, which is closer to true for a futures-based implementation than for a margin account. It does not model the path, and a levered portfolio can be forced to sell at exactly the wrong moment.
What actually happened
Everything above is an expectation. This is the record. The window starts in November 2006 because that is the first month in which every asset class in both portfolios has data, and it runs to August 2026, which is 238 months. It is not a chosen period; a pair holding private equity or inflation-linked bonds simply cannot start earlier than those markets can be measured.
| November 2006 to August 2026 | Swensen (Yale Endowment) | Permanent Portfolio (Browne) |
|---|---|---|
| Annualized return | 5.92% | 6.33% |
| Volatility | 12.87% | 6.91% |
| Sharpe ratio | 0.40 | 0.72 |
| Worst drawdown | -45.58% | -14.04% |
Rebalanced annually, with no fees or trading costs applied, so the figures compare the allocations rather than an implementation of them. Series are index and fund total returns in US dollars, extended backwards before each ETF existed by the longer-running instrument tracking the same exposure.
The record at matched risk
| Both at 12.87% volatility | Exposure | Annualized return |
|---|---|---|
| Swensen (Yale Endowment) | 100% | 5.92% |
| Permanent Portfolio (Browne) | 186% | 9.79% |
One window is one path. A record covering 20 years contains a particular sequence of regimes, and a portfolio that happened to hold the right asset through them will look better than it deserves. Read this next to the forward-looking figures rather than instead of them. Borrowing is charged at the cash rate plus 0.50 percentage points, the same terms as the forward-looking table.
What each one holds
| Asset class | Swensen (Yale Endowment) | Permanent Portfolio (Browne) |
|---|---|---|
| US Intermediate Treasuries | 15% | 25% |
| AC World Equity | n/a | 25% |
| Gold | n/a | 25% |
| Cash / Money Market | n/a | 25% |
| US REITs | 20% | n/a |
| US Large Cap | 15% | n/a |
| EAFE Equity | 15% | n/a |
| Listed Private Equity | 15% | n/a |
| TIPS | 15% | n/a |
| Emerging Markets Equity | 5% | n/a |
The window is unkind to the endowment model
The measurable record here opens in 2006, so it starts a little over a year before the financial crisis. Illiquid and credit-sensitive assets were hit hardest in that episode, and the endowment allocation carries a great deal of both.
Endowments could hold through it because their spending is predictable and their horizon is permanent. A private investor holding the same allocation faced the same fall without those advantages, which is the gap between the model and its reproduction.
Access is the real variable
The expected returns used for private equity and real assets are the published assumptions for those categories, which describe the average fund rather than the funds an endowment of Yale's standing can reach.
Read the endowment figure as the best case, and reduce it for whatever your realistic access costs in fees and quality. The Permanent Portfolio has no equivalent problem, since everything in it can be bought at low cost by anyone.
Frequently asked questions
Is the endowment model better than a simple portfolio?
It carries higher published return assumptions because of its weighting to private and real assets, but those come with illiquidity, wider dispersion between managers and an access problem that most investors cannot solve. Over the period that can be measured, the simpler allocation held up better once risk was equalised.
Why does the comparison start in 2006?
Because that is the first month in which every asset class in both portfolios can be measured with consistent data, and the endowment allocation holds the ones with the shortest histories. The window is set by the data rather than chosen.