Factor-Augmented Regularized Models with Multiple Structural Breaks

王霞，3044am永利集团博士，新加坡管理大学经济学博士后，现任中国人民大学经济学院吴玉章特聘教授、博士生导师，入选国家级青年人才项目。她主要从事理论计量经济学以及宏观经济监测与预测等研究工作，在 International Economic Review, Journal of Econometrics, Journal of Business & Economic Statistics，Econometric Theory、《经济研究》等高水平期刊发表论文三十余篇，已完成的两项国家自然科学基金项目在结题后均获评“特优”。

This paper introduces a factor-augmented regularized model with multiple structural breaks, designed to jointly capture common and idiosyncratic information while accommodating structural instability in high-dimensional dataset. To address the dual challenges of high dimensionality and instability, we propose a two-stage estimation procedure. In the first stage, we combine the split-sample technique with the adaptive group Lasso to achieve consistent variable selection in the presence of instability, while in the second stage, we employ the group fused Lasso to consistently identify both the number and locations of structural breaks. Two information criteria are provided to guide the selection of tuning parameters in the Lasso-type estimations. We derive the asymptotic distributions of the break fractions and the post-Lasso estimators. Monte Carlo simulations demonstrate the excellent finite-sample performance of the proposed estimators. In an application to forecasting the U.S. industrial production growth rate, our method outperforms competing approaches, highlighting the importance of accommodating idiosyncratic information and structural instability into economic models.