|
|
Market Behavior
Project title: Predicting stock price changes.
Project summary: When it comes to stocks, price and value are two distinctly different things. And somewhere along the line the psychology of investors and the markets enters the picture too. This project seeks to determine what several of the major constituents of a stock price are, which is to say: what adds value to a stock price? Working from a multidisciplinary perspective, research and the development of a model focusses on a synthesis of traditional finance theories with recent developments in behavioral research and psychometrics to produce a hybrid approach to choosing stocks systematically. Much of the development work centers on applying behavioral algorythms and sophisticated rating systems, in addition to the refinement of specific predictive factors, to support the model. The model takes into account fundamental corporate data, non-financial/weightless data, and market data. This approach we believe is more robust than present black box systems being developed by many brokerage houses - our understanding is that many of these use data mining approaches which have no underlying rationale for any effects found and often produce only short-lived rewards.
To date, we have produced a functional model that predicts changes in stock prices six months hence with a high degree of statistical accuracy, as tested by multiple regression analysis (R squared = 60.8%). Current work centers on the analysis and forward testing of a larger pool of stocks consisting of about two hundred and fifty stocks that are then whittled down using our filters.
The approach is to an extent value oriented and therefore the stocks being assessed are for medium-term investment of six months, although they can be rolled over for the next period following a re-analysis and selection of the portfolio under consideration. The model is particularly relevant to investors and hedge funds using long-short and market neutral strategies. However, we are also developing a system based on the same model that will allow intraday trading.
Lead investigators: Jonathan Myers, James Fisher.
|