1. ZHAW digitalcollection
2. School of Management and Law
3. Publikationen
4. Publikationen School of Management and Law

Publication type:	Working paper – expertise – study
Title:	Predicting accruals
Authors:	Chardonnens, Patrick
et. al:	No
DOI:	10.2139/ssrn.4309693
Extent:	48
Issue Date:	6-Sep-2023
Publisher / Ed. Institution:	SSRN
Language:	English
Subjects:	Accrual; Supervised machine learning; Earnings management
Subject (DDC):	006: Special computer methods 332: Financial economics
Abstract:	Accruals rapidly revert from one year to another posing a challenge for predicting them one year in advance. Employing the supervised machine-learning models LASSO, random forests, and SVMs on a comprehensive dataset comprising 453 financial variables, we demonstrate the superior predictive accuracy of these models compared to traditional OLS regressions. More importantly, we demonstrate that including numerous explanatory variables in in-sample regressions leads to severe overfitting, which potentially inflates the results' significance. Consequently, we advocate for the adoption of superior out-of-sample methods. Finally, we reveal that accruals and their components have a higher predictive power for next year's accruals than typical Jones models variables. Our research emphasizes the need for further investigation into individual accrual components to gain a deeper understanding of their characteristics, impact, and predictive power on the accumulated accruals.
URI:	https://digitalcollection.zhaw.ch/handle/11475/29780
License (according to publishing contract):	Licence according to publishing contract
Departement:	School of Management and Law
Organisational Unit:	Institute for Financial Management (IFI)
Appears in collections:	Publikationen School of Management and Law

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.