Title: Penalized estimation of sparse high-dimensional single-equation error correction models
Authors: Stephan Smeekes - Maastricht University (Netherlands)
Etienne Wijler - Maastricht University (Netherlands) [presenting]
Abstract: The estimation of high-dimensional single-equation error correction models is considered. We adopt an l1-penalty to obtain sparse parameter estimates and we consider encouraging additional sparsity through the use of a group penalty on the error correction term. Oracle properties are derived under relatively mild assumptions on the error term while allowing the number of parameters to diverge along with the time dimension. Under the strict assumption of Weak Exogeneity (WE) the estimator can be interpreted as an alternative to a well-known test of cointegration. However, without the assumption of WE, our estimation procedure consistently estimates a set of pseudo-parameters for which we obtain analytical expressions. The predictive capability of our method is demonstrated through the use of simulations as well as an empirical application.