We discuss three methodological issues concerning forecasts of the outcome of financial distress. First, we argue that rather than using a binary model the outcome of financial distress should be modeled using a multinomial specification that distinguishes between failure, survival as going concern, and acquisition. We also argue for a random rather than matched-pair sampling technique to better reflect decision making reality. Finally, we investigate the value of using industry-mean adjusted regressors. We find that the binary bankruptcy model is mis-specified relative to the multinomial model, that the matched sampling technique overstates model accuracy and that industry specific intercepts have better explanatory power than industry-adjusted regressors.