Glossary

Structural Breaks

Abrupt shifts in the parameters of a statistical model — for instance, a change in the trend, mean, or variance of a time series. Zivot and Andrews (1992) developed an endogenous structural-break test for unit roots.

Definition

Why It Matters

Structural breaks, if ignored, distort unit-root tests, inflate forecast errors, and produce misleading parameter estimates. A series that appears non-stationary may in fact be stationary around a shifting mean, and failing to account for the break leads to spurious non-rejection of the unit-root null. In emerging markets like Turkey, structural breaks are common due to currency crises, regime changes, and policy shifts. Detecting and modelling breaks correctly is essential for valid inference and reliable forecasting.

Example

The Turkish lira exchange rate appears to follow a random walk when tested with a standard ADF test. However, incorporating a structural break at August 2018 (the currency crisis) using the Zivot-Andrews test rejects the unit-root null, revealing that the series is trend-stationary with a break. Ignoring the break would have led to incorrect first-differencing and loss of long-run information.

Related Terms

Software Notes

SPSS: Not natively available for structural-break tests; use R integration.
R: Use ur.za() from the urca package for the Zivot-Andrews test. The strucchange package provides the Bai-Perron test for multiple breaks and the CUSUM test for recursive residual analysis.
Stata: Use xtbreak for testing and estimating structural breaks. For a single break, use estat sbscusum after regression. The breakpoint command is also available in recent versions.