The “Optimal Portfolio” by Nick Maggiulli
In his analysis of diversification, Nick Maggiulli on Of Dollars And Data mentions an interesting portfolio he built using asset class data from Bullion Vault spanning 1972-2022. The portfolio consists of:
This asset mix would have produced the highest return per unit of risk over 1972–2022 for the asset classes in that dataset, with materially less volatility than the S&P 500 — and it still lost money in roughly one year out of five.
Do not copy this allocation blindly. It is an in-sample optimum: the weights were chosen by looking at the answer, and gold’s 16% share in particular reflects a window that opens in 1972, one year after the end of Bretton Woods and immediately before gold’s greatest run in modern history. Start the window in 1980 instead and the gold weight collapses. And 38% in U.S. housing is not investable as an asset class — for a household it means one leveraged, undiversified, illiquid property, with all the consequences set out in section “Housing’s Role in Optimal Portfolios”. The useful lesson is the final caveat, not the weights: even the ex-post optimal portfolio loses money one year in five, so a losing year tells you nothing about whether your allocation is wrong.