Garbage in, Garbage out: A Theory-Driven Approach to Improve Data Handling in Supervised Machine Learning

Hyde, Steven J.; Bachura, Eric; Harrison, Joseph S.

doi:10.1108/s1479-838720220000014006

Cited by 3 publications

(1 citation statement)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We are cautious of this approach's potential to threaten validity and generalizability by highlighting idiosyncratic linguistic relationships. Hence, we prune features in alignment with the existing literature prior to training the model (Hyde et al, 2023).…”

Section: Deceptionmentioning

confidence: 99%

The tangled webs we weave: Examining the effects of CEO deception on analyst recommendations

Hyde,

Bachura,

Bundy

et al. 2023

Strategic Management Journal

Self Cite

View full text Add to dashboard Cite

Research SummaryOrganizations are punished by analysts and investors when material deceit by their CEO is uncovered. However, few studies examine analysts' responses to deceptive CEOs before their deceit is publicly known. We use machine learning (ML) models to operationalize the likelihood of CEO deception as well as analysts' suspicion of CEO deception on earnings calls. Controlling for analysts' suspicion of deception, we show that analysts are prone to assigning superior recommendations to deceptive CEOs, particularly those deemed as All‐Star analysts. We find that the benefits of CEO deception are lower for habitual deceivers, pointing to diminishing returns of deception. This study contributes to corporate governance research by enhancing our understanding of analysts' reactions to CEO deception prior to public exposure of any fraud or misconduct.Managerial SummaryUndetected deception by CEOs can impact the stock market by influencing analysts' recommendations. Using an advanced ML model, our study measures the likelihood of deception more accurately than previous methods and identifies a tendency among financial analysts to favor deceptive CEOs, particularly high‐status analysts. However, deception is less effective with analysts who are repeatedly exposed to deception. These findings underscore the importance of awareness of potential deception in CEO communications and the need for continuous scrutiny, learning, and adaptability among analysts.

show abstract

Section: Deceptionmentioning

confidence: 99%

The tangled webs we weave: Examining the effects of CEO deception on analyst recommendations

Hyde,

Bachura,

Bundy

et al. 2023

Strategic Management Journal

Self Cite

View full text Add to dashboard Cite

show abstract

Computer Science on Philosophy Perspective

Nasution

2024

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Did Vulture Decline Increase Death Rates in India?

Dalvi,

Ganu,

Kulkarni

et al. 2024

Preprint

View full text Add to dashboard Cite

This commentary critically evaluates the methodology and conclusions of a recent study by Frank and Sudarshan[1] which links the decline of vulture populations in India to an estimated half-million excess human deaths between 2000 and 2005. Utilizing data from India’s Civil and Sample Registration Systems, Frank and Sudarshan attributed the claimed rise in human mortality to the ecological consequences of vulture loss, notably increased feral dog populations and associated rabies, waterborne diseases, and other health hazards. By examining the data reliability as well as independent analysis of the same data, we identify significant limitations in data reliability, potential confounding factors, and the robustness of the study’s statistical assumptions. Our findings question the strength of the patterns themselves and further the soundness of the causal claims, suggesting that improvements in mortality records and many other factors may potentially explain the observed trends. With the current limitations of data, there is no reliable way to estimate quantitatively the human health consequences of vulture decline.

show abstract

Garbage in, Garbage out: A Theory-Driven Approach to Improve Data Handling in Supervised Machine Learning

Cited by 3 publications

References 72 publications

The tangled webs we weave: Examining the effects of CEO deception on analyst recommendations

The tangled webs we weave: Examining the effects of CEO deception on analyst recommendations

Computer Science on Philosophy Perspective

Did Vulture Decline Increase Death Rates in India?

Contact Info

Product

Resources

About