A0341
Title: Estimating high-dimensional Markov-switching VARs
Authors: Kenwin Maung - University of Rochester (United States) [presenting]
Abstract: Maximum likelihood estimation of large Markov-switching vector autoregressions (MS-VARs) can be challenging or infeasible due to parameter proliferation. A sparse framework is adopted to accommodate situations where dimensionality may be of comparable order to or exceeds the sample size. Two penalized maximum likelihood estimators are proposed with the Lasso or the smoothly clipped absolute deviation (SCAD) penalty. It is shown that both estimators are estimation consistent, while the SCAD estimator also selects relevant parameters with probability approaching one. A modified EM algorithm is developed for the case of Gaussian errors, and simulations show that the algorithm exhibits desirable finite sample performance. In applying short-horizon return predictability in the US, a 15-variable 2-state MS-VAR(1) and obtaining the often reported counter-cyclicality in predictability are estimated. The estimators' variable selection property helps identify predictors that contribute strongly to predictability during economic contractions but are otherwise irrelevant in expansions. Furthermore, out-of-sample analyses indicate that large MS-VARs can significantly outperform "hard-to-beat" predictors like the historical average.
