Cointegrated Regressors: Where Folded Concave Penalties Go Beyond LASSO
LASSO fails to consistently estimate cointegrated variables because its standard penalty shrinks them too much. This work introduces Folded Concave Penalties (FCPs) like SCAD and MCP, which are milder than LASSO's L1 penalty. FCPs establish a local oracle property and, when constrained, guarantee a global solution that consistently estimates true coefficients in high-dimensional cointegrated regressions.