Short Selling and Default Prediction: International Evidence

We examine how short selling affects the accuracy of default prediction models. Our results indicate that a dynamic multiperiod logit model accurately predicts 18 percentage points more actual occurrences of default in countries where short-selling is widely practiced. Moreover, our analyses indicate that short selling has virtually no impact on the proportion of inaccurately classified non-default observations. In addition, we find that the lower predictive accuracy in short sale constrained environments is attenuated by the introduction of put option trading. Taken together, these results provide evidence that short selling increases the informativeness of equity market indicators of financial distress without affecting the level of uninformative speculative trading. Finally, in countries that face significant short selling restrictions, our results indicate that greater availability of other, non-market-based, sources of information significantly improves default prediction accuracy, providing evidence of an enhanced role for corporate transparency in settings with significant capital market frictions. _____________________________ We are grateful to the Risk Management Institute for providing the dataset on cross-country default used in this study. Srinivasan acknowledges research support from research grants from the Risk Management Institute at the National University of Singapore. Financial support from the University of Chicago Booth School of Business and the University of Rochester Simon School of Business is gratefully acknowledged. We are grateful for helpful comments received from Bill Beaver, Mike Minnis, Shiva Rajgopal, Wendy Wilson, and workshop participants at Emory University, the University of Rochester, and the 2012 FARS Mid-year Meeting. * Corresponding author. Tel.: +17737029656; [email protected].