The Perks and Perils of Machine Learning in Business and Economic Research

Tom L. Dudda and Lars Hornuf

No 11721, CESifo Working Paper Series from CESifo

Abstract: We examine predictive machine learning studies from 50 top business and economic journals published between 2010 and 2023. We investigate their transparency regarding the predictive performance of machine learning models compared to less complex traditional statistical models that require fewer resources in terms of time and energy. We find that the adoption of machine learning varies by discipline, and is most frequently used in information systems, marketing, and operations research journals. Our analysis also reveals that 28% of studies do not benchmark the predictive performance of machine learning models against traditional statistical models. These studies receive fewer citations, arguably due to a less rigorous analysis. Studies including traditional statistical models as benchmarks typically report high outperformance for the best machine learning model. However, the performance improvement is substantially lower for the average reported machine learning model. We contend that, due to opaque reporting practices, it often remains unclear whether the predictive gains justify the increased costs of more complex models. We advocate for standardized, transparent model reporting that relates predictive gains to the efficiency of machine learning models compared to less-costly traditional statistical models.

Keywords: machine learning; predictive modelling; transparent model reporting (search for similar items in EconPapers)
JEL-codes: C18 C40 C52 (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.cesifo.org/DocDL/cesifo1_wp11721.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:ces:ceswps:_11721

Access Statistics for this paper

More papers in CESifo Working Paper Series from CESifo Contact information at EDIRC.
Bibliographic data for series maintained by Klaus Wohlrabe ().