Main asset classes: long-term behaviour analysis

Damodaran's research shows a clear hierarchy between the asset classes. At one end sits cash, measured by three-month US Treasury bills, which barely beat inflation at 0.32% a year in real terms. At the other sit US small cap equities at 11.98% a year nominal. The table also shows the relationship between risk and return that most investors expect to find. The two highest-returning assets carried the highest volatility and the deepest drawdowns.

One asset refuses to fit. Gold carried almost exactly the volatility of US equities, 19.4% against 19.3%, and delivered 2.49% a year in real terms against 6.77%. Equity risk, bond returns.

Part of that is an artefact of the data. Look at the gold line on the chart below. It is flat until 1971. The United States fixed the gold price at $20.67 an ounce until 1933, then at $35 from the Gold Reserve Act of January 1934 until President Nixon suspended the convertibility of the dollar into gold on 15 August 1971. For 38 of the 98 years in the table, gold had no market price at all. It had an administered one, and inflation ate it. Between 1934 and 1971 gold returned 1.34% a year nominal, which is minus 1.65% a year in real terms, at a volatility of only 6.7%.

Measure gold from 1972 and it becomes a different asset. It returned 8.89% a year nominal against 11.06% for US equities. Its volatility was 27.3% against 17.0%. The correction therefore does not rescue gold. Freed from the peg, gold was considerably more volatile than US equities and still returned less. The chart below traces what a single dollar invested at the start of 1928 became in each asset class, in nominal terms, on a logarithmic scale so that the smaller lines remain readable beside the larger ones.

Readers should be aware of an important limitation of any analysis based on Damodaran's numbers. Damodaran reports calendar year returns, so every drawdown here runs from one year end to another and misses whatever happened inside the year. Gold is the clearest case. The table shows a worst nominal drawdown of 53.5%, from the end of 1980 to the end of 2000. Gold actually peaked at $850 in January 1980 and bottomed near $253 in July 1999. Peak to trough, that is a fall of 70%.

The fact that gold delivered lower returns than equities, with more volatility and an awful twenty-year bear market, does not mean gold should be excluded from a portfolio. An asset can be more volatile than equities and still reduce the volatility and the worst drawdown of the portfolio it joins, provided it does not move in step with what is already there. That is diversification. It is a property of how assets move together, not of any asset on its own. Portfolio construction is a separate subject and this article will not go further into it. But it is the reason that comparing asset classes on return and volatility alone leaves something out. Correlation belongs in the comparison.

I computed the correlation of each asset class with US equities since 1972, using Damodaran's annual returns. A reading of 1 means the two moved in perfect step. A reading of 0 means they moved independently of one another. A reading of minus 1 means they moved in exact opposition. The table below clearly shows that over the last five decades gold had no meaningful correlation with equity returns. Unsurprisingly, US small cap and Baa corporate bonds showed the highest correlation with the broad US equity market.

Correlation of annual nominal returns with the S&P 500, 1972 to 2025, 54 observations, computed from Damodaran's data. Using real returns instead changes little: gold moves to −0.20 and 10-year Treasuries to 0.16. One figure is an average across more than five decades, and the relationship moves a great deal inside them.

Investors rarely put all their eggs in one basket. So how did portfolios spread across several asset classes perform over the long run? I built seven model portfolios and measured them since 1972, against cash and against inflation. Everything is computed from Damodaran's annual returns, with the weights reset to target every year.

The table below shows the performance and risk statistics for 1972 to 2025. Inflation averaged 3.9% a year over that period. The returns shown are nominal, so they do not account for it.

Nominal annual returns, 1972 to 2025, 54 years, computed from Aswath Damodaran's data. Portfolios are rebalanced to their target weights at the end of every year. Drawdowns and time under water are measured on calendar year-end values, so intra-year falls were deeper. Time under water is the longest stretch from a peak until that peak was regained; the fixed income portfolio has not yet regained its 2020 peak. Figures are gross of costs, taxes and any cost of rebalancing. These are illustrative constructions for comparison, not allocations offered to any reader.

These figures come from the Jordà-Schularick-Taylor Macrohistory Database, built by five economists and published in the Quarterly Journal of Economics in 2019 as The Rate of Return on Everything. It is the most complete long-run panel that exists: sixteen countries, every year from 1900, with equities, bonds, bills and housing measured on the same basis. The authors assembled it from national archives rather than from index providers, and the work is peer-reviewed.

It is also a reconstruction, and three limits matter. Housing is the weakest series. Its rental yield is proxied from the rent component of consumer prices rather than measured from net income, and the figures carry no maintenance, taxes, insurance or transaction costs, and assume no mortgage. Germany's bond and bill numbers span two currency destructions, in 1923 and in 1948, which is why its nominal figures mean nothing and its real ones are brutal. And the sample is made of survivors. Every country here kept functioning markets and enforceable property rights. Russia in 1917 and China in 1949 expropriated their investors outright, and neither appears in any dataset of this kind. The real distribution of outcomes facing an investor in 1900 was worse than the one shown here.

Now the findings. The United States had the best equity market of the sixteen, at 6.7% a year after inflation. France and Portugal lost money, at minus 0.2% and minus 0.3%. Seven percentage points a year separate the best from the worst, sustained over 120 years.

The best asset class was not equities. Housing returned a median 6.2% a year real, beat equities in thirteen of the sixteen countries, and was positive in every single one. Equities managed a median of 4.5% and were negative in two. Housing was also steadier across borders: its range is 4.1 points, against 7.0 for equities. Only the United States and Australia saw shares beat property. Read the housing caveat again before drawing conclusions from that, because stripping out ownership costs would narrow the gap considerably.

The worst asset class was cash. Treasury bills returned a median 0.2% a year real and lost purchasing power in seven of the sixteen countries. Bonds did a little better, at 1.4%, and lost money in four. The asset most investors treat as the safe default was, across a century and sixteen countries, barely a store of value.

Measuring how asset classes behave over the long run is harder than it looks. The answer shifts with the country, the period and the measure chosen. Anyone offering a single number has hidden most of the work.

Equities offered the best long-run returns, and charged for them. They led in the United States over a century, and they led again across the seven periods since 1972. But they lost real money over 120 years in two of the sixteen countries studied, and their worst falls reached 98% in three of them. The American record, which almost everyone learns from, had the mildest worst case of the sixteen.

Cash was the safest asset over any short window and the worst over long ones. Treasury bills lost purchasing power in seven of the sixteen countries, and American cash took 64 years to recover its 1933 value. Warren Buffett made the point in his 2011 letter to Berkshire shareholders, writing of currency-denominated investments that “in truth they are among the most dangerous of assets”. Germany shows how far this can go. The bond series returned 5.9% a year nominal over 120 years and minus 18.1% real. Inflation did part of that. The state did the rest. The Bond Redemption Law of July 1925 reinstated some pre-war Reich bonds at 2.5% of face value. The currency reform of June 1948 annulled Reich debt altogether. The London Agreement of 1953 then wrote down what remained of the external debt by about half. Prices ruined the German bondholder once. Statute finished him.

Real estate is where the two main sources used in this article disagree most, and each errs in the opposite direction. The Rate of Return on Everything makes housing the best asset in thirteen of sixteen countries, but proxies rent from consumer prices and charges nothing for maintenance, taxes, insurance or transaction costs. Damodaran measures only house prices and credits no rent at all, which leaves American housing almost six points a year behind equities from 1928 to 2025. The honest answer probably lies between those two, and this article cannot place it precisely.

Gold has traded freely only since 1971. From 1972 it returned 8.9% a year nominal, at a volatility of 27.3% against 17.0% for equities: more risk, less return. Its usefulness was never in its own performance alone. Its correlation with equities sits near zero, and that is what allowed it to lower the risk of portfolios holding it, which is why the balanced portfolio beat the classic 60/40 while owning less equity.

The last finding is the one that resists summary. No two of the seven periods since 1972 rewarded the same thing. Gold's three best stretches were equities' three weakest. Bonds protected investors in every major equity fall except 2022, when they did the opposite. A reader looking for the asset that always works will not find it here, because the record does not contain one. What it contains instead is a method: decide which risk you are least able to survive, and check what each asset did the last time that risk arrived.