Fourier AR Model
Fourier-Augmented Autoregressive Model · Also known as: Fourier AR, trigonometric AR model, smooth transition AR with Fourier terms, FAR model
The Fourier AR model extends the standard autoregressive specification by adding trigonometric (sine and cosine) terms to the deterministic component. This allows the model to capture smooth, gradual shifts in the mean or trend of a time series without requiring the researcher to locate or count structural break points explicitly.
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 the Fourier AR model when you have a univariate time series and suspect gradual regime shifts or smooth trends but cannot pinpoint exact break dates. It is appropriate for medium to long samples (T > 60 is advisable) where standard break-point tests have low power. It is also useful as a robustness check after standard AR estimation. Do not use it when breaks are sharp and abrupt — a dummy-variable or Chow-test approach is more appropriate in that case. Avoid over-parameterising by including too many Fourier frequencies; one or two frequencies are usually sufficient and parsimonious.
Strengths & limitations
- Captures smooth, gradual structural changes without requiring pre-specified break dates or break count.
- Nest the standard AR model as a special case, so standard tests and information criteria can guide frequency selection.
- Estimated by OLS, making it computationally straightforward and widely available in standard software.
- More powerful than break-point tests when true breaks are smooth rather than abrupt.
- Applicable to a wide range of macroeconomic and financial series known for slow regime shifts.
- Assumes breaks are smooth and gradual; performs poorly if the data contain sharp, discrete structural breaks.
- Frequency selection can be data-driven and therefore prone to overfitting if too many frequencies are tried.
- Requires a sufficient sample length (T > 60 recommended) to reliably estimate the Fourier component and the AR lags jointly.
- Does not directly test for the number or timing of breaks — supplementary break-dating procedures are needed for that.
- Interpretation of the fitted Fourier curve can be less intuitive than discrete break dummies.
Frequently asked
How many Fourier frequencies should I include?
Start with k = 1 and test up to k = 5. Select the frequency that minimises the residual sum of squares or an information criterion such as AIC. Using more than two frequencies risks overfitting for typical macro samples.
How is this different from a standard AR model with a structural break dummy?
A dummy variable imposes a sharp, instantaneous break at a fixed date. The Fourier AR model instead approximates smooth, gradual change without fixing a break date, making it more appropriate when regime shifts unfold over several years.
Can I use the Fourier AR model with nonstationary data?
If the series is integrated of order one, you should first apply a Fourier-augmented unit root test (e.g., Enders-Lee 2012 ADF) to determine the integration order, then difference the series before estimating the Fourier AR model in levels of the stationary transformed variable, or embed the Fourier terms inside a cointegration framework.
Does a significant F-test on the Fourier terms confirm a structural break?
Not precisely. A significant F-test indicates that the deterministic component is non-constant — consistent with a smooth break — but it does not identify the number or timing of breaks. Supplementary Bai-Perron or Zivot-Andrews tests are needed to date discrete breaks.
Which software packages support the Fourier AR model?
The model can be estimated in R (using the 'FourierTS' or custom code via the 'lm' function with constructed sine/cosine regressors), Stata (via user-written commands), and EViews. Because it reduces to OLS after the frequency is selected, any software with OLS and flexible regressor construction can implement it.
Sources
- 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 ↗
- 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. DOI: 10.1111/j.1467-9892.2006.00478.x ↗
How to cite this page
ScholarGate. (2026, June 3). Fourier-Augmented Autoregressive Model. ScholarGate. https://scholargate.app/en/econometrics/fourier-ar-model
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.
- ARIMA modelEconometrics↔ compare
- ARMA modelEconometrics↔ compare
- Autoregressive modelEconometrics↔ compare
- Fourier ARDL Bounds TestEconometrics↔ compare
- Fourier VECMEconometrics↔ compare
- Structural Break AR ModelEconometrics↔ compare