A Comparative Study of Logit and Artificial Neural Networks in Predicting Bankruptcy in the Hospitality Industry

Abstract

Taking financial ratios as independent variables, this study used the framework of a neural network applied to hospitality firm bankruptcy, comparing the results to those of a logit model. Based on the empirical results of the two methodologies, the neural network obtained a higher accuracy rate than the logit model in an in-sample test. However, when tested with a holdout sample for verification, both models achieved a 100% accuracy rate. The study found that 'total liabilities to total assets' was a significant variable based on the results of both the t-test and logit analysis. Since hospitality firms are known for being highly leveraged, the conclusion can be drawn that extensive debt financing, when not accompanied by the competitive market value of equity, could play a pivotal role in forcing firms to file for bankruptcy.

Examining Ownership Equity as a Psychological Factor on Tourism Business Failure Forecasting

“…The results of their study showed that the ANN provides a higher accuracy rate for the in-sample test than does the logit model. Park and Hancer (2012) also found that the logit and ANN models both achieved a 100% accuracy rate with a holdout sample.…”

“…However, Kim and Gu (2006) recommended the use of a logit model versus the discriminant model because of its theoretical soundness. Park and Hancer (2012) compared the logit model and ANNs for forecasting bankruptcy within the hospitality industry. The results of their study showed that the ANN provides a higher accuracy rate for the in-sample test than does the logit model.…”

Applying a probabilistic neural network to hotel bankruptcy prediction

“…Results concluded that NN and SVM had an advantage over the others in prediction accuracy, with rates of 92% and 91%, two and three years prior to failure. Park and Hancer (2012) aimed to compare the accuracy of NN to that of a Logit model in predicting bankruptcy in the hospitality industry (hotels, restaurants and entertainment services); so the same sample was used for both analyzed data (40 bankruptcies matched with 40 solvent firms). Results show that NN obtained a higher accuracy rate than Logit in an insample test; and with a hold-out sample, for verification, both models reached a 100% accuracy rate.…”

“…Most of the previous studies related to bankruptcy prediction in the hospitality industry have analyzed one year prior to bankruptcy (Gu & Gao, 2000;Gu, 2002;Kim & Gu, 2006b;Park & Hancer, 2012), and several have analyzed two and three years in advance (Young & Gu, 2010a, Kim & Gu, 2006aLi & Sun, 2012). In this study, three pools of data have been built for the period 2005-2012.…”

“…Although bankruptcy prediction is a common topic in management and finance literature, very few studies refer to the hospitality industry, and only two have focused exclusively on hotels. Studies on predicting bankruptcy in the hospitality industry have used samples consisting of restaurants and hotels (Gu & Gao, 2000;Kim & Gu, 2006a;Park & Hancer, 2012;Pacheco, 2015) and their results do not coincide, as each yields different predictors, possibly because of the heterogeneity of the samples used.…”