Kaddoura's new estimator detects hidden kinks in panel data

A new paper by Yousef Kaddoura, posted on arXiv (2608.07162), tackles a long-standing problem in econometrics: how to detect structural breaks that occur gradually rather than abruptly in panel data. The proposed estimator uses adaptive weighted group penalties applied to second differences of the coefficient path, allowing the model to automatically identify an unknown number of common kink dates across all cross-sectional units. This is the first panel framework that estimates both the number and locations of kinks while accounting for fixed effects, a critical feature when unobserved heterogeneity is present. The theoretical results show consistent recovery of the kink structure, with endpoint slopes converging at the usual cubic regime-length rate and interior slopes at rates determined by their own and adjacent regime lengths.

The paper also extends the method to a coefficient-by-coefficient setting, where each regressor can kink at different dates—a flexible approach for real-world data. Monte Carlo simulations validate the estimator's finite-sample performance, and an application to the debt-growth relationship in macro-finance illustrates its practical value. Written in accessible prose and accompanied by an experimental HTML version, the work bridges econometric theory and machine learning techniques (group lasso, adaptive penalties). For researchers analyzing economic time series, financial panels, or policy evaluation, this method offers a data-driven way to pinpoint exactly when relationships change, without pre-specifying break dates.