On this page
The three books on the Choose page hold 25 to 30% gold. Two decades of data support that weight and two other decades reject it, so the weight is presented as a view, with both records on the table.
The record that supports it
From February 2006 gold returned 10.0% a year, close to stocks’ 11.1%, at the same volatility. From January 2010 it returned 8.2% against stocks’ 14.2%. What made it useful in both windows was 2022, the year stocks fell 18% and Treasuries fell 15% together; gold fell 1%. And in 2008, when stocks fell 37%, gold rose 5%.
A portfolio of stocks and Treasuries has one hedge, and it works when the shock is a growth shock. When the shock is inflation and rates, both halves lose. The Balanced financial independence book of 50% stocks, 20% Treasuries and 30% gold lost 13% in 2022 against 17% for 60/40.
An optimizer given the 2006 window wants 43 to 45% gold and no Treasuries. Given the 2010 window it wants 15 to 25% gold with 70 to 85% stocks. The stock weight swings with the start; the gold weight stays between 15 and 30% in both. That is the one robust thing the two decades say about the mix, and the books hold 25 to 30%.
The record that rejects it
Gold’s London price peaked at $667 in September 1980 and bottomed at $255 in August 1999, a 62% fall over nineteen years. From January 1980 to December 2000 it lost 4.1% a year in dollars and 7.7% a year after inflation. The 1980 peak was regained in April 2007. Over the same twenty years US stocks returned 16.4% a year and ten-year Treasuries 10.6%.
A 30% gold weight held through the 1980s and 1990s cost about three points a year against 60/40, every year, for two decades, with no year in which stocks and bonds fell together to pay it back.
What the two records have in common
Gold earned its place in the 1970s (37% a year from 1971 to 1980, while stocks returned 6.5%, bonds 4.3% and prices rose 7.7% a year), from 2001 to 2011 (20% a year), and since 2006. It lost its place from 1980 to 2000 and from 2011 to 2015, when it fell 35%.
Gold pays in the decade when stocks and bonds both lose to inflation, and costs in the decade when real rates are high and rising. A backtest that starts in 2006 or 2010 contains only the first kind of decade. The Asset allocation page’s rolling-window chart, which starts in 1972, contains both.
What the weight is, then
A 25 to 30% gold weight is a judgment that the next decade has a meaningful chance of looking like 2022 rather than 1995: inflation above target, real rates that cannot rise far because of government debt, and a stock-bond correlation that stays positive. If that judgment is wrong, the cost is two to three points a year for as long as it stays wrong. If it is right, the cost of not holding gold is a year like 2022 with nothing in the book that rose.
That is a view, and the site labels it as one. The evidence label covers the two bad years and the two optima; the view label covers the weight.
Next: 60/40 is the floor.