Size & Regional Distribution of Financial Behavior Patterns in Spain

Zusammenfassung

The relationship between the financial behaviour of firms and some business risk factors have been thoroughly researched during decades. Our paper tries to draw a dynamic two-dimensional map of companies using a broad number of accounting ratios during a period of seven years based on regional and size characteristics. The intention is to explain financial behaviour for a subsequent comparative classification of businesses. For small and medium sized firms, accounting ratios are not just the only data available, but the best suitable building blocks to our research. In order to reduce input dimensionality without losing information, we use two methodologies: the Principal Components Analysis, and correlation analysis. Finally, we also use the whole data set of variables with the intention to see if there are outstanding differences in the final results. Subsequently, the companies are sorted out by size and geographical / political areas or regions, using a neural net model called Self-Organizing Maps. We also use a cluster model as a benchmark. The model allows us to visualize dynamically some traits by means of mapping n variables in two dimensions: We can classify firms in groups, see which variables determine each group, monitor the evolution of firms in terms of group selection, and determine the fittest group in terms of financial ratios and the firms within it. Finally, we also see no outstanding differences between the Principal Component Analysis (PCA) model and the whole data set, whereas the model without high correlation variables shows a weaker stability. Although this analysis represents a step forward for the right valuation of the Spanish small and mid-size firm network, its applicability covers not only more dimensions, but any other country with clear-cut differentiated political regions. The data has been taken from the Spanish Register of Business Names, being the raw material the annual accounting books and records, covering the years 1998- 2004, both included