bibtype J - Journal Article
ARLID 0646076
utime 20260216143905.9
mtime 20260216235959.9
DOI 10.1016/j.jempfin.2026.101705
title (primary) (eng) Deep learning, predictability, and optimal portfolio returns
specification
page_count 22 s.
media_type P
serial
ARLID cav_un_epca*0362207
ISSN 0927-5398
title Journal of Empirical Finance
volume_id 87
publisher
name Elsevier
keyword Return predictability
keyword Portfolio allocation
keyword Machine learning
keyword Recurrent neural networks
keyword Empirical asset pricing
author (primary)
ARLID cav_un_auth*0329534
name1 Babiak
name2 M.
country CZ
author
ARLID cav_un_auth*0242028
name1 Baruník
name2 Jozef
institution UTIA-B
full_dept (cz) Ekonometrie
full_dept Department of Econometrics
department (cz) E
department E
full_dept Department of Econometrics
country CZ
fullinstit Ústav teorie informace a automatizace AV ČR, v. v. i.
source
url https://library.utia.cas.cz/separaty/2026/E/barunik-0646076.pdf
cas_special
project
project_id GA25-18070S
agency GA ČR
country CZ
ARLID cav_un_auth*0484465
abstract (eng) We study the dynamic portfolio selection of an investor who uses deep learning methods to forecast stock market excess returns. In a two-asset allocation problem, deep neural networks — both feedforward and long short-term memory (LSTM) recurrent architectures — deliver economically significant gains in terms of certainty equivalent returns and Sharpe ratios relative to linear predictive regressions. These gains are robust to alternative performance measures, the inclusion of transaction costs, borrowing and short-selling constraints, different rebalancing horizons, and subsample splits, and are particularly pronounced during NBER recessions and periods with large return swings. Within the class of neural networks we consider, economic performance is broadly similar across architectures, with the recurrent LSTM specification providing incremental benefits with more frequent rebalancing. Overall, our evidence suggests that exploiting the time-series structure of standard predictor variables via deep learning can generate meaningful portfolio improvements for investors beyond those obtained from linear models.
result_subspec WOS
RIV AH
FORD0 50000
FORD1 50200
FORD2 50206
reportyear 2027
num_of_auth 2
inst_support RVO:67985556
permalink https://hdl.handle.net/11104/0375846
cooperation
ARLID cav_un_auth*0300496
name Lancaster University
country GB
cooperation
ARLID cav_un_auth*0503333
name Institute of Economic Studies, Charles University, Prague
institution IES, CUNI
country CZ
confidential S
article_num 101705
mrcbC91 C
mrcbT16-e BUSINESS.FINANCE|ECONOMICS
mrcbT16-f 3.3
mrcbT16-g 0.4
mrcbT16-h 9.1
mrcbT16-i 0.00281
mrcbT16-j 0.966
mrcbT16-k 3931
mrcbT16-q 98
mrcbT16-s 0.944
mrcbT16-y 55.38
mrcbT16-x 2.52
mrcbT16-3 781
mrcbT16-4 Q1
mrcbT16-5 2.300
mrcbT16-6 79
mrcbT16-7 Q2
mrcbT16-C 70
mrcbT16-M 0.76
mrcbT16-N Q2
mrcbT16-P 74.1
arlyear 2026
mrcbU14 SCOPUS
mrcbU24 PUBMED
mrcbU34 WOS
mrcbU63 cav_un_epca*0362207 Journal of Empirical Finance 0927-5398 1879-1727 Roč. 87 č. 1 2026 Elsevier