Bayesianische Portfolio Regularisierung
Zusammenfassung der Projektergebnisse
This project has demonstrated the improvements of computationally intensive Bayesian methods and machine learning when applied to high-dimensional portfolio choice problems. The classical portfolio weight shrinkage methods were reformulated into a regression framework, which allowed for application of Bayesian regularization methods. Moreover, in the project a generalization of asymptotic theory for the newly developed statistical learning technique, sparse approximate factor models, was developed and shown to outperform the existing approaches. Furthermore, the project contributed to the literature by providing and explanation for the puzzling empirical findings of the naive equally weighted portfolio. Portfolio evaluation procedures commonly used in the literature has been shown to have low power and thus the the null hypothesis of equal performance of the equally weighted portfolio compared to many theoretically superior strategies cannot be rejected in many out-of-sample horse races. Finally, the drawbacks of the considered methods were shown to be mitigated with the help of machine learning tools applied to the high-dimensional portfolio choice. This project has contributed to a better understanding of the statistical properties of highdimensional portfolio models and provided the literature with extensive empirical comparisons of the proposed methods to the commonly used approaches.
Projektbezogene Publikationen (Auswahl)
- “Bayesian Shrinkage of Portfolio Weights,” Working Paper. Available at SSRN
Frey, Christoph and Pohlmeier, Winfried
(Siehe online unter https://dx.doi.org/10.2139/ssrn.2730475) - (2017): Three Essays on Bayesian Shrinkage Methods, Ph.d. thesis, University of Konstanz, Konstanz
Frey, C.
- (2019): Three Essays on Covariance Matrix Estimation and Factor Models in High Dimensions, Ph.D. thesis, University of Konstanz, Konstanz
Zagidullina, A.
- (2019): Three Essays on Robust Inference in Economics and Finance, Ph.D. thesis, University of Konstanz, Konstanz
Kazak, E.
- (2019): “Testing out-of-sample Portfolio Performance,” International Journal of Forecasting, 35 (2), 540–554
Kazak, E., & Pohlmeier, W.
(Siehe online unter https://doi.org/10.1016/j.ijforecast.2018.09.010) - (2019): “Widened Learning of Index Tracking Portfolios,” 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA), Boca Raton, FL, USA, pp. 1800–1805
Gavriushina, I., Sampson, O., Berthold, M. R., Pohlmeier, W., & Borgelt, C.
(Siehe online unter https://doi.org/10.1109/ICMLA.2019.00291) - “Sequential Stock Return Prediction Through Copulas,” DEA Working Papers 91, Universitat de les Illes Balears, Departament d'Economía Aplicada
Audrone Virbickaite & Christoph Frey & Demian N. Macedo
- (2020): Sparse Approximate Factor Estimation for High-Dimensional Covariance Matrices
Daniele, M., Pohlmeier, W., & Zagidullina, A.
- (2020): Three Essays on Regularization Methods in High-Dimensional Factor Models, Ph.D. thesis, University of Konstanz, Konstanz
Daniele, M.
- Bayesian Shrinkage of Portfolio Weights
Frey, C., & Pohlmeier, W.
(Siehe online unter https://dx.doi.org/10.2139/ssrn.2730475) - “The CAPM with Measurement Error: There?s life in the old dog yet!” Jahrbücher für Nationalökonomie und Statistik - Journal of Economics and Statistics, 240 (4), 417–453
Simmet, A., & Pohlmeier, W.
(Siehe online unter https://doi.org/10.1515/jbnst-2018-0089) - (2020): “Valid Inference for Treatment Effect Parameters under Irregular Identification and Many Extreme Propensity Scores,” Journal of Econometrics
Heiler, P., & Kazak, E.
(Siehe online unter https://doi.org/10.1016/j.jeconom.2020.03.025)
