Agenda

Management lecture - Fundamentals, Analysts and and Portfolio Selection: a Machine Learning Approach

Speaker: prof. Stefano Bonini, Stevens Institute of Technology, School of Business

Abstract:
A large body of literature has tried to identify the relative contribution of market returns, companies fundamentals, analysts recommendation and the content of analysts report,  jointly and in isolation, to returns. Yet results are mixed. Portfolios performance is highly sensitive to the relative weights of these factor and their  contribution to performance varies over time. On a very large-scale sample of over 1.5 million sell-side analysts reports, covering almost 1,700 stocks over 25 years, we implement a machine learning approach and show how managers can use the available information to dynamically form forward-looking portfolios. We obtain a host of important results. First, portfolios formed using time-varying loadings of information inputs generate a strongly significant 8% abnormal return over the sample period. Second, the relative contribution of inputs exhibits substantial time series variation. Third, analysts-related information  contribution to portfolios formation is large. Fourth, the textual content of reports (net tone) is significantly more valuable than outright recommendations. Overall, our analysis outlines a viable approach to incorporate multiple information inputs in portfolio managers decisions.

join the zoom meeting https://unive.zoom.us/j/85835976629
Meeting ID: 858 3597 6629
Passcode: 3q230f