Structural Break GLS
Generalized Least Squares with Structural Breaks · Also known as: GLS with structural breaks, break-adjusted GLS, structural change GLS, regime-switching GLS
Structural Break GLS combines Generalized Least Squares estimation with explicit allowance for regime shifts in the data-generating process. The method estimates separate coefficient vectors for each segment defined by detected break dates while correcting for non-spherical errors — heteroscedasticity or autocorrelation — that frequently accompany structural change, yielding consistent and efficient estimates across all regimes.
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 Structural Break GLS when you have time-series or panel data where the regression relationship plausibly changed at one or more points in time and where standard OLS residuals display heteroscedasticity or autocorrelation. Typical applications include long macroeconomic series spanning policy regimes, financial return series affected by market crises, and energy or commodity price regressions covering supply shocks. Prefer this approach over plain structural break OLS when Breusch-Pagan or Breusch-Godfrey tests reject homoscedasticity or no-autocorrelation. Do not use it when there are no signs of structural instability (stable Chow test) or when the sample is too short to estimate separate parameters reliably in each sub-period; a minimum of around 30 observations per regime is a practical lower bound.
Strengths & limitations
- Simultaneously handles structural instability and non-spherical errors, combining two important corrections in one estimator.
- BLUE under correct specification of both break dates and error covariance, yielding efficient inference.
- Bai-Perron global search procedure selects the number and location of breaks objectively, reducing researcher discretion.
- Produces segment-specific coefficient estimates that reveal how relationships evolve across economic regimes.
- More reliable standard errors than OLS in the presence of heteroscedastic or autocorrelated errors around break dates.
- Efficiency gains depend heavily on correctly specifying the error covariance structure; a misspecified Omega can make FGLS worse than OLS.
- Estimating break dates introduces pre-testing uncertainty; inference conditional on estimated breaks is approximate.
- Requires sufficient observations in each regime; with many breaks or short series, segment-specific estimates become imprecise.
- Computationally more demanding than OLS, particularly when the number of candidate break dates is large.
Frequently asked
How does Structural Break GLS differ from plain structural break OLS?
Both allow coefficients to change at estimated break dates, but GLS additionally corrects for non-spherical errors by weighting observations with the inverse of the estimated error covariance. When errors are heteroscedastic or autocorrelated — which is common around structural breaks — GLS produces more efficient estimates and valid standard errors, whereas OLS standard errors are biased in those circumstances.
How many breaks should I allow?
Use the Bai-Perron sequential or global information-criterion procedure to determine the number of breaks objectively. As a rule of thumb, each segment should contain at least 30 observations (or 15 % of the sample) to support reliable estimation of the segment-specific coefficient vector.
What if I do not know the error covariance structure?
Estimate Omega from OLS residuals obtained in a first pass, using a regime-specific variance estimator or a Newey-West kernel estimator for autocorrelated errors. This yields a Feasible GLS (FGLS) estimator that is asymptotically as efficient as infeasible GLS when Omega is consistently estimated.
Can I apply this method to panel data?
Yes. In a panel setting, structural breaks can be common across units or unit-specific. Panel GLS with structural breaks extends the time-series framework by pooling information across units, and the Panel Bai-Perron tests formalise detection of common break dates.
Is Structural Break GLS robust to outliers?
Not inherently. GLS down-weights high-variance observations but does not explicitly trim outliers. If outliers drive apparent break detections, consider combining break testing with robust GLS (M-estimator-based) or pre-screening for influential points.
Sources
- Bai, J., & Perron, P. (1998). Estimating and testing linear models with multiple structural changes. Econometrica, 66(1), 47–78. DOI: 10.2307/2998540 ↗
- Greene, W. H. (2012). Econometric Analysis (7th ed.). Prentice Hall. ISBN: 978-0131395381
How to cite this page
ScholarGate. (2026, June 3). Generalized Least Squares with Structural Breaks. ScholarGate. https://scholargate.app/en/econometrics/structural-break-gls
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.
- Generalized Least SquaresStatistics↔ compare
- Panel GLSEconometrics↔ compare
- Robust GLSEconometrics↔ compare
- Structural Break OLSEconometrics↔ compare
- Structural Break WLSEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare