Interpretable machine learning for earnings forecasts: Leveraging high-dimensional financial statement data

Dieter Hess, Frederik Simon and Sebastian Weibels

No 25-06, CFR Working Papers from University of Cologne, Centre for Financial Research (CFR)

Abstract: We predict earnings for forecast horizons of up to five years by using the entire set of Compustat financial statement data as input and providing it to state-of-the-art machine learning models capable of approximating arbitrary functional forms. Our approach improves prediction one year ahead by an average of 11% compared to the traditional linear approach that performs best. This superior performance is consistent across a variety of evaluation metrics as well as different firm subsamples and translates into more profitable investment strategies. Extensive model interpretation reveals that income statement variables, especially different definitions of earnings, are by far the most important predictors. Conversely, we find that while income statement variables decline in relevance, balance sheet information becomes more significant as the forecast horizon extends. Lastly, we show that the influence of interactions and non- linearities on the machine learning forecast is modest, but substantial differences between firm subsamples exist.

Keywords: Earnings Forecasts; Cross-Sectional Earnings Models; Machine Learning (search for similar items in EconPapers)
JEL-codes: G11 G12 G17 G31 G32 M40 M41 (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.econstor.eu/bitstream/10419/323935/1/1932922806.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:zbw:cfrwps:323935

Access Statistics for this paper

More papers in CFR Working Papers from University of Cologne, Centre for Financial Research (CFR) Contact information at EDIRC.
Bibliographic data for series maintained by ZBW - Leibniz Information Centre for Economics ().