All Weather and the Permanent Portfolio: four-season allocations tested with ETFs

All Weather is the idea of mixing assets so that one account can hold up whether the economy speeds up or slows down and whether inflation rises or falls. Ray Dalio's Bridgewater started running it in 1996, and it became widely known after Tony Robbins printed a simple set of weights that individuals could copy in his 2014 book Money: Master the Game. This article runs those public weights and Harry Browne's Permanent Portfolio with ETFs and compares them with 60/40 and 100% SPY. The figures are as of September 30, 2026 and were calculated by the operator.

Rules

• All Weather (the public simplified version): US stocks 30%, long-term US Treasuries 40%, intermediate Treasuries 15%, gold 7.5% and commodities 7.5%. These are the weights Robbins's book printed under the name All Seasons, based on his interview with Dalio. They are not the actual holdings of the Bridgewater fund. The ETFs are SPY, TLT (maturities over 20 years), IEF (7 to 10 years), GLD and DBC. The book names the whole US stock market for stocks; this article uses SPY to match the site's other articles.
• Permanent Portfolio: 25% each in stocks, long Treasuries, gold and cash. In his 1999 book Fail-Safe Investing, Browne explained it as picking one strong asset for each of four conditions: prosperity, inflation, deflation, and tight money or recession. Instead of an ETF, cash earned the US 3-month Treasury bill rate (^IRX) every day. With the short Treasury ETF SHY as cash, it returned 7.1% a year with a max drawdown of −18.4%.
• Comparisons: 60/40 is SPY 60% and IEF 40%, the same mix as in the Asset allocation article. 100% SPY was run under the same conditions.
• Conditions: to match DBC, the last of the five ETFs to list (February 2006), $10,000 went in at the open on March 1, 2006, with nothing added later. Weights were reset to target at the open of the first trading day of each January, and monthly resets were run separately. A 0.15% one-way cost was charged on every traded amount, dividends and interest were reinvested, and everything was sold at the close on September 30, 2026.

Results (as of September 2026)

The worst year is the lowest calendar-year return from 2007 to 2025. The starting year 2006 and 2026, which runs only to September, are left out.

All Weather's volatility, measured from daily returns, was 8.0% a year, less than half of SPY's 19.3%. It lost money in three of the 19 years from 2007 to 2025: 2015 (−3.0%), 2018 (−2.9%) and 2022 (−18.8%). In exchange its annual return was lower, and after 20 years and 7 months its balance was about 43% of SPY's. The Permanent Portfolio returned 0.4 points a year more than All Weather with a max drawdown 6.1 points smaller. Its 25% in cash softened 2022, when rates rose, and its 25% in gold contributed heavily in 2025, when GLD rose 63.7% (the Permanent Portfolio gained 22.4% that year).

2008 and 2022

In 2008, when SPY lost 36.8%, TLT gained 34.0% and IEF 17.9%. All Weather, with 55% in Treasuries, ended the year up 3.3%, and its deepest drop within the year was only −14.2% (November 12, 2008). 2022 was the opposite. As the 3-month Treasury bill rate rose from 0.05% at the start of the year to 4.26% at the end, SPY (−18.2%), TLT (−31.2%) and IEF (−15.2%) all fell, and only DBC, the commodity ETF, gained (+19.3%). All Weather lost 18.8%, more than SPY. It fell 23.0% from its high on December 27, 2021 to October 20, 2022, and regained that high on July 1, 2025, 3.5 years later. SPY regained its January 3, 2022 high on December 13, 2023.

The difference came down to whether stocks and Treasuries moved in the same direction. The correlation of daily SPY and TLT returns was −0.42 from March 2006 to 2021, and it was negative in every single year from 2007 to 2021. From 2022 to September 2026 it was +0.11, and it was positive in every year from 2022. All Weather's weights rest on the premise that Treasuries rise when stocks fall, so once that relationship changed, the 55% in Treasuries did not cushion the decline and made it deeper.

Looking at risk instead of weights

Bridgewater describes All Weather as starting from equal risk across four cases, growth above or below expectations and inflation above or below expectations, with leverage applied to low-volatility bonds so that their impact matches stocks (The All Weather Story). The simplified version aims for a similar effect without leverage by raising the weight of volatile long Treasuries to 40%.

Treating the weights as fixed and splitting the simplified version's daily variance by source, from March 2006 to 2021 it came from SPY 39%, TLT 40%, IEF 6%, gold 7.5% and commodities 7%. The 30% in stocks and the 40% in long Treasuries carried almost equal risk. From 2022 to September 2026, the two Treasury funds' share grew from 46% to 58%, and portfolio volatility rose from 7.6% to 10.1% a year. As interest rates became more volatile, Treasuries turned from an asset that reduced risk into one that added to it.

The Bridgewater fund's own yearly results are not public. Reports and posts put its 2022 loss at around −20%, but the figures differ by source and could not be confirmed. One public record exists: the State Street Bridgewater All Weather ETF (ALLW), listed on March 5, 2025. According to the issuer, its one-year NAV return to August 31, 2026 was 17.71%, below its benchmark MSCI ACWI IMI (22.35%). As of September 30, 2026, the same page lists exposures of 70.78% to global nominal bonds, 43.52% to global equities, 41.32% to inflation-linked bonds and 32.64% to commodities, 188% in total. Unlike the simplified version, its use of leverage shows in the numbers.

Rebalancing frequency

Resetting the weights on the first trading day of every month gave All Weather 6.3% a year (max drawdown −23.7%), the Permanent Portfolio 6.8% (−17.5%) and 60/40 8.2% (−32.6%), all slightly below yearly resets. All Weather traded 110 times with yearly resets and 1,240 times with monthly ones, paying $133 and $256 in costs. 2008 showed a large gap. All Weather reset yearly gained 3.3%; reset monthly, it returned −0.01%. Monthly resets sold rising Treasuries to buy falling stocks twelve times that year.

Browne wrote that the portfolio should be checked once a year and reset only when an asset has moved below 15% or above 35%. Run that way, the Permanent Portfolio was reset in four years, 2010, 2015, 2022 and 2025, and returned 7.2% a year with −16.9%. Left alone after the first purchase, All Weather returned 7.2% a year with −24.6%, but by September 2026 its stock share had grown from 30% to 62%, making it a different account from its name.

What research and public records show

• The calculation in Robbins's book: the book states that these weights returned 9.72% a year net of fees from 1984 to 2013, lost money in four of 30 years, and had a largest loss of −3.93% in 2008. This article's ETF calculation gives +3.3% for 2008, the opposite sign. The book used indexes for the years before ETFs existed, and the 2008 result depends heavily on which commodity and Treasury indexes are used (DBC lost 31.8% in 2008). It is also worth noting that long-term US interest rates mostly fell from 1984 to 2013.
• Bridgewater: the firm's article says All Weather started in 1996 with Dalio's family trust assets and was not first designed as a product. The fund's yearly results could not be confirmed from public sources.
• The stock-bond correlation: in a 2023 Journal of Portfolio Management paper, AQR's Brixton, Ilmanen and coauthors found that the stock-bond correlation was negative for the two decades after 2000 but mostly positive in the 1970s through the 1990s, and that it moves toward positive when uncertainty about expected inflation is high. The post-2022 figures above point the same way.
• Individual assets: over the same period TLT returned 2.5% a year and fell 48.4% from its high on August 4, 2020 to October 19, 2023. DBC returned 2.7% a year and fell 76.4% from its July 2008 high to April 2020. GLD returned 9.7% a year, second only to SPY. Much of All Weather's result came from what each asset earned in this period rather than from how they were mixed.

When the rule worked and when it did not

• It worked in years when only stocks fell hard: in 2008 All Weather gained 3.3% and the Permanent Portfolio 0.9%, while SPY lost 36.8%. In the March 2020 crash, All Weather's drop of −13.6% was also shallower than SPY's −33.7%.
• It failed in a year of fast-rising rates: in 2022 All Weather lost 18.8%, more than SPY, and took 3.5 years to regain its high, longer than SPY (1.9 years).
• It earns less in a rising stock market: $10,000 put in during March 2006 became $86,318 in SPY and $37,107 in All Weather.
• Risk is smaller but not evenly small: it lost money in only three years from 2007 to 2025, but the 2022 loss alone was more than three times the other two combined.

Related concepts and search terms

• Risk parity: splitting a portfolio so each asset carries equal risk rather than equal weight. Leverage on low-volatility bonds is the usual way to get there, and Bridgewater's All Weather is the best-known example.
• All Seasons portfolio: the name of the simplified version in Robbins's book. In Korea it is mostly called the All Weather portfolio.
• Permanent Portfolio: Browne's four assets at 25% each. A US mutual fund named Permanent Portfolio Fund also exists, but its weights differ and it is not the same thing.
• 60/40 portfolio: the reference mix of 60% stocks and 40% bonds. More figures are in the Asset allocation article.
• Rebalancing bands: resetting only when a weight moves outside a set range. Browne's 15 to 35% rule is one example. Results of holding weights while investing every month are in the stock and cash mix backtest.
• Compared with income: using dividend stocks instead of bonds or gold to reduce swings is covered in Dividend growth investing with SCHD. Dividend stocks are still stocks and cannot play the role of Treasuries in a year like 2008.

Search terms: all weather portfolio, all seasons portfolio, permanent portfolio, risk parity, Ray Dalio asset allocation, stock bond correlation 2022, and the Korean terms for the All Weather portfolio, the four-season portfolio and the permanent portfolio.

Limits of these numbers

• The calculation covers one period, March 2006 to September 2026. The first 15 years were a time of falling rates when Treasuries rose, and the last 5 a time of rising rates. Different start and end dates could change the ranking.
• This is a simplified version built from ETFs, not the Bridgewater fund's result. The fund uses leverage, inflation-linked bonds and non-US assets.
• DBC holds commodities through futures, so gains and losses from rolling futures are mixed into its return, and its fee is higher than the other ETFs'. GLD pays no dividends, and its fee is taken from the gold it holds. Both funds' costs are already in the prices.
• Cash interest uses the 3-month Treasury bill rate as is. Real deposits and short-term bond ETFs differ, and Korean investors face exchange rates and taxes on top. For Korean residents, gains on foreign ETFs above ₩2.5 million a year are taxed at 22% (including local income tax).