Cycle No. 02 · Empirical · contested
Kondratiev wave
~50–60 year economic long wave: technological expansion → plateau → stagnation → trough. Kondratiev traced it in price, interest-rate, and production series reaching back to the 1780s
Kondratiev wave · Does it hold up?
Not testable on the record that exists — and that is the finding, not a dodge.
The record this verdict is judged on, US TFP growth (annual, unsmoothed), runs 77 years: 1.4 of the 3.0 full periods this site requires before it will run a test on a 54-year claim. Roughly 85 more years of that measurement would reach the floor.
This is not a verdict about this theory in particular: 0 of the 9 paired constructions on this site clear that floor. Long-cycle claims are hard to test because the records are short, not because the theorists are careless.
By this page's curve, 2026 sits at a peak (the curve tops out in 2027; cos +0.99). The next low falls around 2054. That is a position of this construction, not the theorist's forecast. Every cycle in 2026 →
- Period
- 54 years
- Reference peak
- 1973
- Paired data
- TFP growth · 5-yr
Peak calibration
Move the peak yourself: calibrate this cycle →
Conventional anchor inside the post-WWII K-wave's 1968–1974 turning-point window (oil shock + Bretton Woods end). Other revivalists prefer 1968 (Mandel) or 1971; we use 1973 as the most cited single year.
CaveatThe periodicity claim has published renunciations - Metz, who once reported K-wave evidence, later withdrew it, and Solomou's long-run studies find no regular wave. Spectral detections (e.g. Korotayev & Tsirel 2010) remain contested. The wave is kept here as the most-cited long-wave theory, not as settled empirics.
Confidence classification: Empirical · contested — an empirical literature exists on the claim, and it is disputed. It grades the theorist's own evidence for the period; a paired data series is a separate comparison this site added. Every cycle on this site is a pure sinusoid built from the theory's stated period and one documented reference peak — a deliberately naïve construction, so that disagreement between theories rather than parameter fitting is what you see. See about for why nearly every cycle peaks near the present.
Where this curve plots its extrema
Computed from period 54 and reference peak 1973, across the site's default window (1600–2050). These are positions of this construction, not dates claimed by the theorist.
- Peaks
- 1649 · 1703 · 1757 · 1811 · 1865 · 1919 · 1973 · 2027
- Troughs
- 1622 · 1676 · 1730 · 1784 · 1838 · 1892 · 1946 · 2000
Paired data series
US TFP growth (5-yr rolling)
% per year5-year centered rolling average of Fernald's utilization-adjusted US TFP growth, derived by this project from the annual `dtfp_util` column. Annual data begins 1948; the build script keeps clipped (asymmetric) windows at the boundaries rather than dropping rows, so the 1948 and 1949 endpoints are edge artifacts
Why this pairingKondratiev waves predict 50–60 year cycles of technological paradigm expansion and exhaustion. TFP growth is the most direct measurable output.
- Source
- Fernald (2014), FRBSF Working Paper 2012-19
- License
- freely available; © FRBSF, no explicit reuse license
Spectral verdict
Insufficient data — no test possible · 1.4 of 3.0 required periods
The US TFP growth (annual, unsmoothed) record spans 77 years — 1.4 of the 3.0 full periods this site requires before testing a 54-year claim. No test was run and no p-value exists: the honest verdict is that the record is too short to test the claim at all.
Why two labels: the “Paired data” line and section on this page, and the citation at the end, name the 5-year rolling average of this series, which is the version the interactive chart draws. The site measures the record's length on the annual figures, and would run any test on them, because a rolling average makes neighbouring years move together and would distort the background-noise model a test compares against.
“Insufficient data” is an eligibility outcome under the site's pre-registered 3.0-period rule - the test is declined, not failed - and is not evidence for or against the theory.
Target period 54 years · Insufficient data — no test possible · 1.4 of 3.0 required periods
Swipe sideways for the whole figure
Scroll sideways for the whole figure, or click it to open it full size
No test was run: this record covers 1.4 of the 3.0 full periods the pre-registered test requires, so the spectrum in the figure is descriptive only. Had the record qualified, the test would be a harmonic regression at the exact stated period against an AR(1) red-noise null (99,999 bootstrap draws). Read the protocol under spectral testing on the methods page, or fetch the machine-readable verdicts.json.
Reuse this
Kooi, D. (2026). Sinusoidal History: Kondratiev wave. https://sinusoidalhistory.com/cycles/kondratiev · DOI 10.5281/zenodo.21998618. Paired data series: US TFP growth (5-yr rolling) (Fernald (2014), FRBSF Working Paper 2012-19), freely available; © FRBSF, no explicit reuse license.
Our code and the original chart files are MIT-licensed as our own work. The plotted data stay under their upstream terms, listed above: CC BY series carry their attribution requirement with them, and a series with no explicit license is shown here without any grant we can pass on. Reuse our figures and code with attribution; honour the data terms for the plotted series.
Open this cycle in the interactive chart to calibrate its peak and period against the data, or compare it with the other nine on one axis.