http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2589272

Predicting Large Negative Stock Returns: The Trouble Score

1. Korcan AkUniversity of California, Berkeley – Haas School of Business
   April 6, 2015

Abstract: 
The purpose of this study is to predict large negative stock returns. I create an indicator variable, which takes the value of one when a company experiences a stock price decline of 50 percent or more over the subsequent year. Then, using both accounting-based and market-based variables, I predict this large change by using a logit model; the outcome of this prediction is the “Trouble Score”. The top three deciles of the T-Score correctly classify 63.00 percent of observations for the in-sample tests and 60.52 percent of observations for the out-of-sample period. Further, I document a significant and negative association between the out-of-sample T-Score and future stock returns even after controlling for well-known risk factors and control variables that are correlated with future stock returns.