Predicting SMEs failure: logistic regression vs artificial neural network models

Abstract: Research questions: This study compares the power of logit and
artificial neural network (ANN) models in predicting the failure of SMEs in the
hospitality industry and identifies the predictors that are significant in
determining business failure. Motivation: SMEs are an important segment of
the Malaysian economy and contribute significantly to the country’s economic
growth. However, SMEs are riskier and associated with a high failure rate. In
Malaysia, around 3.5% of the SMEs in the hospitality industry fail within the
first two years and 54% of them cease operations within four. Idea: The use of
ANN to model business failure, particularly in the hospitality industry, is
relatively unexplored in the emerging markets. Based on the literature, this
study hypothesizes that ANN models outperform logit models because of less
stringent model assumptions. Data: Excluding missing information, a matched
sample of 41 failed and 41 non-failed SMEs in the hospitality industry was
identified from the year 2000 to 2016. The accounting ratios, firm-specific
characteristics and governance variables are selected as potential predictors of
SMEs failure in the hospitality industry. Method/Tools: Stepwise logit
regression and multilayer perceptron ANN models were used to determine
significant predictors to predict business failure. Each model’s predictive
power was compared. Findings: The ANN model was found to consistently
outperform the logit model in classifying the failed and non-failed SMEs in the
hospitality industry. Furthermore, the ANN model ranked liquidity as the most
important predictor, followed by profitability and leverage, in determining
business failure. Board size was also found to be a significant predictor in
addition to the financial variables. The stepwise logit model also suggests a
significant relationship between board size and the failure of SMEs. Therefore,
in addition to financial predictors, a firm’s governance is also key to business
survival. Contributions: The findings of this study contribute to the limited
literature on SMEs in the hospitality industry by providing empirical evidence
from an emerging market perspective. The failure prediction model can be
utilized to warn of potential business failure so that strategic measures can be
taken to mitigate the risk of failure.