Buffalo, NY – Assistant Professor of Finance Marc LoGrasso, PhD, is exploring how artificial intelligence can inform investing.
Can AI pick dependable stocks for everyday investors
In his latest research, he tested whether ChatGPT-4 could select stock portfolios from 1985 to 2021 using only information available at each point in time—a rigorous experiment that spans decades of market booms, busts, and crises. His study, published in the journal Modern Finance, found that ChatGPT-generated portfolios consistently delivered positive returns, averaging about 1% per month over two-year holding periods, and never significantly underperformed the market - even during major downturns like the 1987 crash, the dot-com bubble, and the 2008 financial crisis.
The inspiration for LoGrasso’s research came from a seemingly unlikely source: online investment articles. About two years ago, he read pieces by Thomas Yeung at InvestorPlace suggesting that ChatGPT could be used to select stocks.
“I started thinking about people who want to invest in the stock market but don’t know where to begin and wouldn’t go to a financial advisor for various reasons,” LoGrasso said. “These individuals might turn to a large language model like ChatGPT for guidance. The question became: would that actually work?”
This curiosity led him to test ChatGPT’s ability to generate stock portfolios over nearly four decades using only information available at the time—creating a comprehensive test of AI stock-picking.
ChatGPT's top stock picks
LoGrasso’s research shows that ChatGPT follows a cautious, top-down strategy, favoring large, stable companies in sectors such as technology and healthcare. Household names like Microsoft, Apple, and Johnson & Johnson appeared repeatedly in the AI’s portfolios. The study found that ChatGPT’s approach emphasizes stability over aggressive outperformance, making it a practical reference point for risk-averse, unsophisticated investors seeking diversified exposure to proven growth sectors.
“This study shows both the promise and limits of AI as a financial advisor,” said LoGrasso. “ChatGPT may not replace human expertise or sophisticated portfolio optimization, but it demonstrates how emerging technologies can help everyday investors make informed decisions while minimizing downside risk.”
Shawn O’Rourke, PhD, dean of the Division of Business, Communication and Health Studies, highlighted the impact for students and alumni.
“Marc’s research connects directly to the challenges and opportunities facing today’s investors. It gives our students and graduates evidence-based insights into AI in finance and showcases Canisius University’s commitment to preparing them for a rapidly evolving marketplace.”
LoGrasso’s study, “Could ChatGPT Have Earned Abnormal Returns? A Retrospective Test from the U.S. Stock Market,” is available open access at Modern Finance.
Canisius was founded in 1870 in Buffalo, NY, and is one of 27 Jesuit colleges and universities in the U.S. Consistently ranked among the top institutions in the Northeast, Canisius offers undergraduate, graduate and pre-professional programs distinguished by close student-faculty collaboration, mentoring and an emphasis on ethical, purpose-driven leadership.
