Search

The importance of micro-entities due to their generation of employment and propelling economic activity, together with the fact of their particularities, implies the need to design appropriate methods that anticipate their bankruptcies. For that purpose, a hybrid model by combining parametric and nonparametric approaches is developed in this paper. First, the variables with the highest predictive power to detect bankruptcy are selected using logistic regression (LR). Subsequently, a non-parametric method, namely regression trees and classification (CART), is then applied to companies classified as "bankruptcy" or "non-bankruptcy". Our results show that this model provides a better result than when it is implemented in isolation, which joins its easier interpretation and faster convergence. Moreover, we demonstrate that the introduction of non-financial and macroeconomic variables complement the financial ratios for bankruptcy prediction. Findings are based on a data set of micro-entities (MEs), as recently defined by the European Union.

La importancia de las micro-entities como generadoras de empleo y propulsoras de la actividad económica conlleva, unida a sus mayores tasas de quiebra y a su dificultad para acceder a las fuentes de financiación, la necesidad de diseñar métodos apropiados que anticipen sus quiebras. Con este fin, en este trabajo se desarrolla un modelo híbrido mediante la combinación de enfoques paramétricos y no paramétricos para la detección de sus quiebras. Para ello, se seleccionan las variables con mayor poder predictivo para detectar la quiebra mediante un modelo híbrido de regresión logística (LR) y árboles de regresión y clasificación (CART). Nuestros resultados muestran que este modelo híbrido obtiene una mejor performance que aquellos modelos implementados de forma aislada, además de tener una más fácil interpretación y una convergencia más rápida. Por otra parte, se constata la conveniencia de la introducción de variables no financieras y macroeconómicas que complementen a la información proporcionada por los ratios financieros para la predicción de la quiebra de las micro-entities, lo cual está en línea con las características propias e idiosincrasia de este tamaño empresarial recientemente definido por la Comisión Europea.------------------------------------The importance of micro-entities due to their generation of employment and propelling economic activity, together with the fact of their particularities, implies the need to design appropriate methods that anticipate their bankruptcies. For that purpose, a hybrid model by combining parametric and nonparametric approaches is developed in this paper. First, the variables with the highest predictive power to detect bankruptcy are selected using logistic regression (LR). Subsequently, a non-parametric method, namely regression trees and classification (CART), is then applied to companies classified as "bankruptcy" or "non-bankruptcy". Our results show that this model provides a better result than when it is implemented in isolation, which joins its easier interpretation and faster convergence. Moreover, we demonstrate that the introduction of non-financial and macroeconomic variables complement the financial ratios for bankruptcy prediction. Findings are based on a data set of micro-entities (MEs), as recently defined by the European Union., Artículo revisado por pares