Fourier OLS (Fourier-Augmented Ordinary Least Squares)
Fourier-Augmented Ordinary Least Squares · Also known as: Fourier OLS, Fourier-augmented OLS, trigonometric OLS, smooth structural break OLS
Fourier OLS is an OLS regression extended by adding low-frequency trigonometric (sine and cosine) terms to the regressor matrix. These Fourier components approximate smooth, gradual structural changes in the regression relationship over time without requiring knowledge of the number, timing, or form of the breaks.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Fourier OLS when you have time-series or panel data with a continuous outcome and you suspect that the regression relationship has changed gradually over the sample period — but you do not know when or how many breaks occurred. It is well-suited to macroeconomic and financial data where policy changes, structural reforms, or business cycles create smooth rather than abrupt shifts. Do not use it as a substitute for well-specified economic regressors: the Fourier terms capture residual smooth variation but cannot replace theoretically motivated controls. Avoid it when genuine sharp breaks are expected (use Chow-type structural break tests instead), when the sample is very short (T < 40), or when interpretability of every parameter is paramount, since the Fourier coefficients themselves have no direct economic interpretation.
Strengths & limitations
- Accommodates smooth, gradual structural change without requiring the researcher to specify the number, timing, or functional form of the break.
- Keeps the closed-form OLS estimator, so estimation is fast and straightforward even with Fourier augmentation.
- Low-frequency Fourier terms are parsimonious: one or two trigonometric pairs often capture the dominant pattern of smooth change.
- Compatible with standard post-estimation diagnostics (residual tests, HAC standard errors, information criteria) already familiar to OLS users.
- Flexible across many OLS extensions: can be combined with GLS, WLS, fixed effects, and ARDL frameworks.
- The Fourier terms absorb smooth variation but do not identify the economic source of structural change — they improve fit without explaining what drove the change.
- Frequency selection via information criteria can be unreliable in short samples (T < 50), leading to over- or under-fitting.
- When multiple sharp breaks coexist with smooth change, Fourier terms alone may not adequately capture the true DGP.
- Adding Fourier regressors increases the regressor count, which can inflate standard errors and reduce power in small samples.
- Standard OLS inference is valid only if residuals are well-behaved; Fourier augmentation does not by itself correct for autocorrelation or heteroscedasticity.
Frequently asked
How do I choose the Fourier frequency k?
Estimate the model for each integer k from 1 up to some maximum (commonly T/2 or 5) and select the k that minimises AIC or BIC. In practice, k = 1 or 2 captures most economically relevant smooth breaks.
Is Fourier OLS the same as a Fourier unit root test?
No. Fourier OLS is a regression model that uses Fourier terms to capture smooth structural change in the level relationship between variables. Fourier unit root tests (e.g., FADF, FKPSS) embed Fourier terms inside a unit root regression and use specialised critical values — they answer a different question.
Can I combine Fourier augmentation with robust or HAC standard errors?
Yes. The Fourier terms are just additional regressors, so all standard post-estimation corrections — heteroscedasticity-robust, HAC, and cluster-robust standard errors — apply exactly as in ordinary OLS.
How many Fourier terms should I include?
Typically one frequency (k = 1) with its sine and cosine pair suffices. Adding more frequencies improves approximation but consumes degrees of freedom. Information criteria penalise model complexity, so they naturally prevent over-fitting.
What if the structural change is sharp rather than smooth?
Fourier terms approximate smooth, gradual shifts poorly when the true break is abrupt. In that case, use dummy-variable Chow tests, the Bai-Perron multiple break procedure, or the Zivot-Andrews test, which are designed for sharp structural breaks.
Sources
- Becker, R., Enders, W., & Hurn, S. (2004). A general test for time dependence in parameters. Journal of Applied Econometrics, 19(7), 899–906. DOI: 10.1002/jae.751 ↗
- 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. DOI: 10.1111/j.1468-0084.2011.00662.x ↗
How to cite this page
ScholarGate. (2026, June 3). Fourier-Augmented Ordinary Least Squares. ScholarGate. https://scholargate.app/en/econometrics/fourier-ols
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Fourier ARDL Bounds TestEconometrics↔ compare
- Fourier Granger CausalityEconometrics↔ compare
- Nonlinear OLSEconometrics↔ compare
- OLS RegressionEconometrics↔ compare
- Structural Break OLSEconometrics↔ compare
- Time-varying parameter OLSEconometrics↔ compare