Fourier ARMA Model
Fourier-Augmented Autoregressive Moving Average Model · Also known as: Fourier ARMA, ARMA with Fourier terms, trigonometric ARMA, smooth structural change ARMA
The Fourier ARMA model augments the classical Autoregressive Moving Average framework with low-frequency Fourier (sine and cosine) terms to capture smooth, gradual shifts in the mean or trend of a time series. Unlike dummy-variable approaches, it requires no prior knowledge of when structural change occurred, approximating change with flexible trigonometric functions.
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 ARMA model when a univariate time series appears stationary but shows a slowly drifting mean or mild trend that cannot be attributed to a single, identifiable break date. It is well-suited to macroeconomic or financial series spanning long samples where institutional or structural change was gradual — for example, inflation regimes, interest rates, or commodity prices across business cycles. Prefer it over break-dummy ARMA when you cannot identify or justify a precise break date. Do not use it if the series is non-stationary (difference first, then consider ARIMA-Fourier), or if the apparent drift is sharp and localized (a Zivot-Andrews break test is more appropriate). Also avoid adding many Fourier harmonics in short samples, as the model quickly becomes over-parameterised.
Strengths & limitations
- Accommodates smooth, gradual structural change without specifying a break date.
- Parsimonious: one or two Fourier pairs typically suffice, adding only two to four parameters.
- Robust to misspecification of the change point; the trigonometric approximation is flexible.
- Combines interpretable ARMA dynamics with a principled trend component.
- Easily estimated by standard ML or NLS routines available in most econometric software.
- Not appropriate for non-stationary series; the series must be made stationary before applying the model.
- Fourier terms model smooth, global change and may miss abrupt or localised structural breaks.
- Frequency selection (the number of harmonics n) requires care; information criteria provide guidance but not certainty.
- May be unfamiliar to referees or policy audiences accustomed to standard ARMA or regime-switching models.
Frequently asked
How do I choose the number of Fourier frequencies?
Estimate the model for n = 1, 2, and 3 frequencies and select the specification that minimises AIC or BIC. In most applied work n = 1 captures the dominant smooth change, and adding higher harmonics rarely improves fit enough to justify the extra parameters.
Is a Fourier ARMA the same as a seasonal ARMA?
No. Seasonal ARMA (SARIMA) handles periodic cycles that repeat at fixed, short intervals (monthly or quarterly). Fourier ARMA uses very low frequencies to capture a single long, smooth arch or drift across the entire sample — typically representing decades of gradual structural change rather than repeating seasonality.
What if my series is non-stationary?
First difference the series to achieve stationarity (or use a Fourier ARIMA model that includes differencing), then apply the Fourier ARMA to the differenced series. Applying the model directly to an I(1) series will produce spurious results.
Can I test whether the Fourier terms are necessary?
Yes. Perform an F-test or likelihood ratio test on the joint significance of the sine and cosine coefficients. If they are not significant, the plain ARMA model is preferred and the Fourier extension adds no value.
How does this differ from the structural break ARMA model?
The structural break ARMA adds dummy variables at known or estimated break dates, producing a sharp level shift. The Fourier ARMA instead bends the mean continuously using sinusoidal functions, better matching gradual transitions. Choose Fourier when the change is smooth and the break date is uncertain; choose the break-dummy approach when there is a clear, abrupt event.
Sources
- Becker, R., Enders, W., & Hurn, S. (2006). A general test for time dependence in parameters. Journal of Applied Econometrics, 21(7), 1005–1028. link ↗
- Enders, W., & Jones, P. (2016). Grain prices, oil prices, and multiple smooth breaks in a VAR. Studies in Nonlinear Dynamics and Econometrics, 20(4), 399–419. DOI: 10.1515/snde-2014-0101 ↗
How to cite this page
ScholarGate. (2026, June 3). Fourier-Augmented Autoregressive Moving Average Model. ScholarGate. https://scholargate.app/en/econometrics/fourier-arma-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
- Fourier ARDL Bounds TestEconometrics↔ compare
- Fourier VAR modelEconometrics↔ compare
- Nonlinear ARMA modelEconometrics↔ compare