The wheel does turn, but only in the commodity price itself.
Relative return over the past 36 months predicts the next 12 months with a coefficient of −0.20 and a t-statistic of −7.62 across 5,130 observations. The same signal on sector equity indices gives only −0.05, which is not statistically significant. Neutralising the Jakarta Composite does not change it.
Traded naively, the hypothesis loses
Buying the two sectors with the most depressed commodities returned −2.8% a year against a benchmark of +7.1%. That −9.3% gap is the only backtest result in the study that is statistically significant, and its sign is negative.
Three reasons explain it. Equity beta to its own commodity runs only 0.26 to 0.90, with R² of 6% to 16%. For all seven sectors, R² against the Jakarta Composite is higher, at 18% to 41%. And below a 22-month horizon, momentum dominates rather than reversion.
Downswings and upswings are the same length
Across 181 datable phases since 1960, the median downswing and the median upswing both run 22 months. Half of all downswings last 15 to 35 months, upswings 14 to 34 months. The short 2010-onward sample suggests downswings drag on much longer, and the full history does not support it.
What is asymmetric is magnitude, not duration. Declines run 35% to 43%, advances 59% to 134%. Missing an upswing costs more than sitting through a downswing.
One strategy out of seven survives
We tested seven mechanical rules over January 2013 to June 2026, then put each through five robustness tests: a parameter grid, a subsample split, dropping one sector at a time, a one-month execution delay, and transaction costs.
Buying the two cheapest commodities returned −2.8% a year, trailing the benchmark by 9.3%, with a t-statistic of −2.43.
Buying the two strongest commodities returned +13.2% a year, ahead by 5.7%, t-statistic 1.31. It passed all five robustness tests.
Buying the deepest equity drawdowns with a momentum confirmation returned +12.1%, ahead by 4.7%, t-statistic 0.74. It passed four of five.
Buying the cheapest commodities with a momentum confirmation returned +10.8%, ahead by 3.5%, t-statistic 0.61. It passed only two of five, and turns negative once costs reach 0.50% per side.
The benchmark, seven sectors equal weighted, returned +7.1% a year. The Jakarta Composite returned +2.3% over the same window.
Only buying the strongest commodities passed all five, which is, ironically, the opposite of the hypothesis under test.
Even that edge is unproven. It wins in just 52% of rolling 12-month windows, once trailed the benchmark for 23 straight months, and drew down 59.2%, deeper than the benchmark itself.
Where things stand in June 2026
Measured against its own real price history since 1960, only nickel and palm oil are genuinely cheap. Nickel sits at the 33rd percentile and palm oil at the 39th, both below their long-run medians despite rising 18.2% and 12.8% respectively in the first half of 2026.
The middle of the pack is oil at the 69th percentile, up 36.2% in six months, and coal at the 72nd, up 28.6%.
At the expensive end sit tin at the 92nd percentile, up 28.7%, copper at the 94th, up 15.0%, and gold at the 99th, the only one to fall, by 1.9%. Gold now trades 336% above its long-run median.
Six of seven commodities rose in the first half of 2026, yet only one sector equity index has positive six-month momentum. Copper rose 15% while its equities fell 38%.
Three takeaways
One. Stop using "it has fallen a lot" as a reason to buy. It is the only significant finding here, and its sign is negative.
Two. Use cycle duration to calibrate patience, not to time entries.
Three. Stock selection matters more than sector selection. Within coal alone, beta ranges from 0.01 to 0.46.
Data. World Bank Pink Sheet, release of 2 July 2026, covering January 1960 to June 2026, deflated by US CPI. Equity prices from Yahoo Finance adjusted close for 44 Indonesian listings, January 2010 to July 2026. Log returns, equal-weight seven-sector benchmark, Newey-West standard errors.
Limitations. Pink Sheet frequency, a monthly average, and equity prices, a month-end close, are misaligned, so reported R² is a lower bound. Realigned, average R² rises from 10% to 19%. Thirteen of 43 listings show stale pricing. The universe carries survivorship bias.
Quantitative research, not investment advice. The full 21-page report is available for download https://drive.google.com/file/d/1xDtg0QeZrOMng2wGKmdK5G5lrR6hwn0o/view?usp=sharing