Date of Award

2016

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

McDonald, Patrick

Area of Concentration

Applied Mathematics

Abstract

Occasionally company's statutory earnings announcements are coupled with very large drops in a company's stock price. I address whether such drops can be foretold by prior earnings statements. Because such price behavior can be attributed to market discovery of earnings statement mismanagement, Beneish, using financial statement ratios, proposed a model to identify potential earnings manipulations by comparing these against earnings restatements in subsequent disclosures. I utilize machine learning techniques to further enhance this score. In addition, instead of earnings restatements, I target stock price drops of 15% and 25% at the next earnings announcement. I then compare the performance of the original Beneish M-Score against the new enhanced score when both are used as a trade signal. I find that this novel model enhances trading performance, with particularly strong improvements for large capitalization stocks.

Share

COinS