The usefulness of the double entry constraint for predicting earnings

Research output: Contribution to journalArticle

Authors

External organisations

  • Kingston University

Abstract

In the absence of an income statement, earnings can be calculated as cash flow from operating activities (CFO) plus accruals, rather than being stated as the difference between income statement revenues and expenses. Following the study by Christodoulou and McLeay (Contemp Account Res 31:609:328. https://doi.org/10.1111/1911-3846.12038, 2014), this paper uses a system of structural regressions with a framework of two simultaneous linear models, allowing the most basic property of accounting—double entry bookkeeping—to be incorporated as a constraint. The paper aims to investigate whether the constrained seemingly unrelated regression (SUR) estimator with two simultaneous models, produces lower out-of-sample prediction errors than each standalone model. We also examine if CFO and accruals are more capable of predicting future earnings than income statement earnings and expenses. Our findings show that in predicting earnings: (1) a system of structural regressions with two constrained simultaneous models produces significantly smaller out-of-sample prediction errors than each separate regression; and (2) accruals and CFO produce smaller out-of-sample prediction errors than earnings and expenses.

Bibliographic note

Khansalar, E. & Kashefi-Pour, E. Rev Quant Finan Acc (2018). https://doi.org/10.1007/s11156-018-00783-3

Details

Original languageEnglish
Number of pages17
JournalReview of Quantitative Finance and Accounting
Early online date11 Dec 2018
Publication statusE-pub ahead of print - 11 Dec 2018

Keywords

  • double entry constraint, accruals, earnings prediction, seemingly unrelated regression