Key points
- The 42 commodities' waves rose and fell at the same times. In 500 tries with random prices, that never happened.
- It still holds when inflation is removed in a different way.
- The four waves found here fall at the same times as those found by other researchers.
- But where we are in a wave cannot be told at the time; it becomes clear years later.
How the waves are extracted
Each real price is split by speed into three parts that add back up to the price: the long-run trend (movements over 70 years), the wave or supercycle (20–70 years), and shorter swings (under 20 years, such as ordinary booms and recessions). The split uses a standard statistical filter (Christiano–Fitzgerald), the same method as Cuddington & Jerrett (2008), Erten & Ocampo (2013) and Jacks (2019).
TEST 1
Do the commodities move together?
A tool that pulls out waves will find them in any series, even pure noise. What it cannot do is make unrelated series rise and fall at the same time. So the test is whether the waves of different commodities move together more than chance allows.
The fake prices keep each commodity’s own year-to-year changes, including their size and long-run drift, but shuffle their order so the commodities are independent of each other. We built 500 sets of 42 fake commodities and ran the same filter.
If the waves were noise, the share would hover around half, inside the grey band. Instead it swings between nearly all and almost none, and it does so at the same times as the index waves.
TEST 2
Is it just how prices are adjusted for inflation?
Real prices are nominal prices divided by the same price index (the deflator), so a swing in that index moves every commodity at once and could create the co-movement by itself. We repeat the test for 1960–2025 with the same 24 commodities deflated two ways: by US consumer prices (Jacks) and by the World Bank’s index of manufactured export prices (MUV), which Erten & Ocampo prefer.
| Divided by | Real data: how closely they move (0–1) | Fake prices: 95 in 100 below | Fake prices: highest |
|---|---|---|---|
| US CPI (Jacks) | 0.73 | 0.04 | 0.15 |
| Manufactures unit value, MUV (World Bank) | 0.59 | 0.05 | 0.12 |
The co-movement survives a different deflator. US CPI does have its own 20–70-year wave (a typical swing of 12% over 1947–2024), so part of the real-price waves is a mirror of inflation; but it is not what makes the commodities move together.
TEST 3
Do the dates match the literature?
| Study | Waves found | Source |
|---|---|---|
| This site (Jacks data, 1900–) | 1903→1917, 1932→1947, 1966→1981, 1996→2011 (trough→peak) | — |
| Jacks (2026 chartbook) | Three complete cycles; the latest began in 1996, peaked in 2011 and is "now likely near its trough". | Jacks, D.S. (2026), "Chartbook of Real Commodity Prices, 1850-2025" |
| Erten & Ocampo (2013), non-oil | 1894–1932 (peak 1917), 1932–1971 (peak 1951), 1971–1999 (peak 1973), 2000s upswing. Waves of 30–40 years, 20–40% above or below trend. | Erten, B. & Ocampo, J.A. (2013), "Super Cycles of Commodity Prices Since the Mid-Nineteenth Century." World Development 44 (UN working paper version) |
| Cuddington & Jerrett (2008), metals | Three supercycles in about 150 years, and the early phase of a fourth. | Cuddington, J. & Jerrett, D. (2008), "Super Cycles in Real Metals Prices?" IMF Staff Papers 55(4) |
| Heap (2005, Citigroup) | Upswings of 10–35 years, complete cycles of 20–70 years, driven by the industrialisation of a major economy. | Heap, A. (2005), "China: The Engine of a Commodities Super Cycle." Citigroup (as summarised in Erten & Ocampo 2013) |
Different data, weights and deflators produce the same broad chronology: peaks around the First World War, after the Second World War, in the 1970s and around 2010, with troughs in the 1930s, 1960s and late 1990s.
What the tests do not show
- A filter can make waves out of noise
Pull a band of frequencies out of a random series and you get plausible-looking waves (the Slutsky effect). That is why this site tests whether commodities move together, not whether waves exist; a filter cannot make independent series move in step.
- There are only four waves to count
125 years since 1900 contain three completed waves. Any regularity such as "upswings last about 15 years" rests on four examples.
- The 20–70-year band is a convention
We use the band the literature uses. A different band moves the peak and trough years by a few years.
- The last few years get rewritten
The filter is unreliable at the end of the sample because it cannot see what comes next. Where we are now is only settled about a decade later (see the Now page).
- The yardstick changes the shape
Jacks's index uses 1975 production weights, so energy weighs heavily, which is why the 1970s wave looks so large. An index weighted towards farm goods emphasises other waves.
Updated 29 Sep 2026 02:33 JST · annual data to 2025 · monthly prices to Aug 2026