Fourier KPSS Test for Stationarity with Smooth Structural Breaks
Also known as: Fourier KPSS, flexible Fourier stationarity test, F-KPSS, KPSS with Fourier approximation
The Fourier KPSS test extends the standard KPSS stationarity test by embedding a flexible Fourier series in the deterministic component of the model. This approach captures smooth, gradual structural breaks in the level or trend of a time series without requiring the researcher to specify the number or timing of those breaks, yielding more reliable inference under structural change.
Key highlights
- Captures smooth, gradual structural breaks without pre-specifying their number or timing.
- Controls size better than the standard KPSS test when the true process has a slowly shifting mean or trend.
- Flexible: the optimal Fourier frequency is data-determined, requiring minimal researcher discretion.
- Provides a formal test of stationarity (null) rather than a test of a unit root, complementing ADF-family tests.
- Applicable to a wide range of macroeconomic, financial, and environmental time series.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the Fourier KPSS test when you suspect a time series is stationary but its mean or trend has shifted gradually over time — for example, macroeconomic aggregates that transition slowly between regimes or climate series with smooth long-run changes. It is preferable to the standard KPSS when structural breaks are suspected but neither their number nor their timing is known. Do not use it as a substitute for sharp-break tests (Zivot-Andrews, Lee-Strazicich) when breaks are abrupt, or when the sample is very short (fewer than 50 observations), as the Fourier terms consume degrees of freedom and critical values are based on asymptotic theory.
Strengths & limitations
- Captures smooth, gradual structural breaks without pre-specifying their number or timing.
- Controls size better than the standard KPSS test when the true process has a slowly shifting mean or trend.
- Flexible: the optimal Fourier frequency is data-determined, requiring minimal researcher discretion.
- Provides a formal test of stationarity (null) rather than a test of a unit root, complementing ADF-family tests.
- Applicable to a wide range of macroeconomic, financial, and environmental time series.
- The Fourier approximation works best for smooth, low-frequency breaks; abrupt level shifts are better handled by sharp-break unit-root tests.
- Requires a moderately large sample for asymptotic critical values to be reliable; small samples inflate size distortion.
- Adding Fourier terms reduces degrees of freedom and can reduce power if breaks are absent.
- Critical values depend on the number of Fourier terms included, requiring care in tabulation or simulation.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How does the Fourier KPSS differ from the standard KPSS test?
The standard KPSS regresses the series on only an intercept (or intercept plus trend). The Fourier KPSS adds sine and cosine terms to that regression, allowing the deterministic component to absorb smooth structural shifts before the stationarity statistic is formed. This prevents the test from confusing a gradual level change with a unit root.
What is the null hypothesis of the Fourier KPSS test?
The null hypothesis is that the series is stationary (possibly around a smooth structural break captured by the Fourier terms). The alternative is that the series contains a unit root. This is the opposite of the ADF or Fourier ADF null, so the two tests are usefully complementary.
How do I choose the Fourier frequency k?
The standard approach is to estimate the Fourier regression for k = 1, 2, ..., K_max (typically K_max = 5) and select the frequency that minimises the sum of squared residuals. Values of k = 1 or 2 are most common in applied work, as they represent low-frequency (smooth) shifts rather than high-frequency oscillations.
Can the Fourier KPSS handle multiple structural breaks?
Yes, in principle. Including multiple Fourier frequencies (k = 1 and k = 2, for example) allows the model to approximate more than one smooth shift. However, each additional frequency uses up degrees of freedom and may reduce power if the breaks are not actually smooth.
Should I use the Fourier KPSS or the Fourier ADF test?
Ideally both, because they test opposite nulls. The Fourier ADF tests the null of a unit root; the Fourier KPSS tests the null of stationarity. Consistency across both tests — failing to reject stationarity in KPSS and rejecting the unit root in ADF — gives the strongest evidence that the series is stationary around a smooth trend.
Sources
- 1.Becker, R., Enders, W., & Lee, J. (2006). A stationarity test in the presence of an unknown number of smooth breaks. Journal of Time Series Analysis, 27(3), 381-409.
- 2.Enders, W., & Lee, J. (2012). A unit root test using a Fourier series to approximate smooth breaks. Oxford Bulletin of Economics and Statistics, 74(4), 574-599.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Fourier KPSS test. ScholarGate. https://scholargate.app/econometrics/fourier-kpss-test