Search

El objetivo de este artículo es la obtención de sendos modelos de predicción del fracaso empresarial en una muestra emparejada y otra aleatoria de pequeñas y medianas empresas con domicilio en Castilla y León (España), a fin de determinar si el poder predictivo de los modelos elaborados está afectado por el método utilizado para seleccionar la muestra objeto de cada estudio. Para ello, consideramos como variables independientes un conjunto de ratios financieros, que reducimos a partir de la aplicación previa de un análisis de componentes principales. Mediante regresión logística, identificamos los factores que mejor predicen el fracaso en ambas muestras, observándose diferencias no solo en las variables significativas, sino también en los resultados de clasificación, lo que conforma la influencia del método de muestreo en los modelos. || This paper focuses on the development of both failure prediction models on a paired sample and a random sample of small and medium-sized firms with head offices located in the region of Castilla y León (Spain), in order to prove if the predictive power of the developed models is affected by the method used to derive the sample aim of each study. To estimate both models, we consider a set of financial ratios as independent variables in each one, which is first reduced by the application of a principal components analysis. Next, a logistic regression analysis is applied to identify those variables that best explain and predict failure in the two samples, where differences in the significant variables and the classification results are observed, which confirms the influence of the sampling method on the business failure prediction results.

This paper focuses on a cross-industry analysis of business failure, in order to identify the explanatory and predictor factors of this event that are different in three of the main industries in every economy: manufacturing, building and service. For each one of these industries, the same procedure is followed. First, a principal components analysis is applied in order to identify the explanatory factors of business failure in the three industries. Next, these factors are considered as independent variables in a discriminant analysis, so as to predict the firms' failure, using not only financial information expressed by ratios, but also other non-financial variables related to the firms, as well as external information that reflects macroeconomic conditions under which they develop their activity. [ABSTRACT FROM AUTHOR]

Purpose – The purpose of this paper is to focus on the development of a business failure prediction model on a sample of small and medium-sized firms with head offices located in the region of Castilla y León (Spain), in order to prove the significance of non-financial information on the prediction of business failure. Design/methodology/approach – In order to reach the authors' aim, one of the most used predictive statistical methods in this field (logistic regression) is applied, in which the authors consider financial ratios and non-financial information as potential variables to predict failure. But before developing the respective models, in order to reduce the number of variables, a principal components analysis (PCA) is first applied. Then, the achieved results with this analysis are used in the prediction step, so as to estimate the models. Findings – The results of the predictive method show that non-financial information, which becomes significant in the developed models, helps financial ratios to improve the ability to predict failure, so any business failure model should also consider both types of information to be accurate. Originality/value – Most of the developed business failure prediction models have used a paired sample with the same number of failed and non-failed firms, which has the drawback of not being representative of the population from which it is chosen. In order to obtain a representative sample, a random sampling method is applied, on the basis of the population size and composition. The selected sample assures that parameter estimates are not inconsistent and biased, as the statistical methods assume. [ABSTRACT FROM AUTHOR]