Financial Determinants of Hotel Bankruptcy in Greece: A Multi-Period Logit Estimation

The aim of this study is to research the financial determinants of bankruptcies for firms belonging exclusively to the economic activity 'hotels and similar accommodation'. The study adopts the stance, that estimations and models focused on specific sectors can be a more precise assessment of firm failure in contrast to generic models that sample firms from broader sectors. The dataset covers the years 2010-2020 and uses available financial and bankruptcy information from two credible data sources for over 5,500 hotels which includes 13 bankruptcy events. A multi-period logistic regression model with clustered robust standard errors is applied, a very competent econometric methodology not applied before in hotel failure studies in this exact form. This type of setting is considered a close-to-reality econometric experiment examining the whole economic activity. A stepwise procedure led to three influential financial ratios with the main results revealing that the likelihood of hotel bankruptcy is positively related to leverage and size in terms of the natural logarithm of total assets and negatively related to EBITDA to total liabilities. By extending the bankruptcy horizon further by one and two years, information content appears fairly robust and all variables retain their sign, as leverage remains stable across all horizons, EBITDA to liabilities remain significant for the one additional period, while sizes' statistical significance becomes intermittent. Comparisons are performed also in terms of sensitivity and variable relevance with two earlier Greek studies for the whole economy and the hotel sector, and in both cases sensitivity is reduced and some variables are rendered not relevant. Implications stemming from the empirical results, suggest that leverage is the most influential factor increasing the probability of bankruptcy, so liabilities should be continuously controlled. Size in terms of asset acquisition should be mitigated too, as the 'too big to fail' principle seems not to hold. Also, a balance of liabilities coverage is recommended. The comparisons in terms of sensitivity, variable relevance and the opposite sign for size implies that bankruptcy modeling per economic activity is more appropriate, stressing the need for an ad hoc variable selection that includes size also, as a predefined set of variables may not always be the most optimal set.