Surviving Chapter 11 Bankruptcies: Duration and Payoff?

Abstract

Three of the authors previously developed a model to predict the duration of Chapter 11 bankruptcy and the payoff to shareholders (Partington et al ., 2001). This work augments that study using a much larger sample to reestimate the model and assess its stability. It also provides an opportunity for out-of-sample testing of predictive accuracy. The resulting models are based on Cox's proportional hazards model and the current article points to the need to test two important assumptions underlying the model. First, that the hazards are proportional and, second, that censoring is independent of the event studied. Using the extended data set, all the previously significant accounting variables drop out of the model and only two covariates of the original model remain significant. These are the market wide credit spread and the market capitalization of the firm, both measured immediately prior to the firm's entry to Chapter 11. Receiver operating characteristic curves are then used to assess the predictive accuracy of the original and extended models. The results show that Lachenbruch tests can provide a misleading indication of predictive ability out of sample. Using the Lachenbruch method of in-sample testing, both models show predictive power, but in a true out-of-sample test they fail dismally. The lessons of this work are relevant to better predicting the gains and losses likely to accrue to shareholders of companies in Chapter 11 bankruptcy and in similar administrative arrangements in other jurisdictions.

Data system for assessing probability of failure in SME reorganization

“…Partington et al (2001) showed that significant firm-specific covariates in reorganization success prediction are the ratio of EBITD to total assets (profitability), the log of market value (size) and the current ratio (liquidity). Wong et al (2007) showed that the ratio of EBITD to total assets (profitability) and the log of market value (size) are statistically significant covariates. Fisher (2007) found that current ratio and the ratio of current secured and preferred claims to current liabilities were statistically significant financial covariates.…”

Dynamic forecasts of financial distress of Australian firms

“…The ROC curve plots the combinations of the false alarm rate (x-axis) and the hit rate (y-axis) as the cut-off point is varied across all possible values. A detailed explanation of the ROC curve can be found in Hanley and McNeil (1982), Mason and Graham (1999), Sobehart and Keenan (2001), and Wong et al (2007).…”

“…Therefore, survival analysis is a natural choice for bankruptcy prediction since it allows the estimation of the probability that a firm survives or goes bankrupt at each point in time t over the forecast period, given the 'random nature of the lifetime' of a company (Peat, 2007: 303). From 2000 onwards there has been growing use of survival analysis in financial distress modelling (Bonfim, 2009;Campbell et al, 2008;Chava and Jarrow, 2004;Cole and Wu, 2009;Nam et al, 2008;Partington et al, 2001;Peat, 2007;Shumway, 2001;Wong et al, 2007).…”

Discussion of Wong et al.

“…This is explained by the fact that my data were randomly selected from the population of firms undergoing financial reorganization in Canada. In contrast, many studies of bankruptcy—including Wong et al . (2007)—deliberately over‐sample firms with publicly held stock.…”

“…The size of the firms in the data is quite small: Average assets are roughly $2 million 2 . Thus, firms in the sample are SMEs, as opposed to the type of firms analysed in Wong et al . (2007).…”