Modelling Credit Risk for Innovative SMEs: the Role of Innovation Measures

Both the PATSTAT and BULLETIN database are available to any user under the request of the EPO. The data have been managed by with SQL and STATA software toolboxes. For more details on this task see Thoma et al.(2010).

Commission Recommendation 96/280/EC of April 3, 1996, updated in 2003/361/EC of May 6, 2003. See http://​europa.​eu/​scadplus/​leg/​en/​lvb/​n26026.​htm.

This implies for the intermediary a reduction in capital requirement proportional to the firm’s size The reduction applies to the capital function through the correlation, which is reduced by a maximum of 0.04 for the smallest firms. This correction is justified by the assumption that defaults of small firms are less correlated and therefore less risky on the whole for the portfolio.

The turnover is given by the sum of sales and net stocks of the reference year. In the present analysis, we use turnover and not sales because AMADEUS does not report information on sales for some countries such as the UK, Ireland, and Denmark.

Linear probability models are not normally used in this literature given that the estimated coefficients may imply probabilities outside the unit interval. For a discussion of the appropriateness of the logit model for default prediction studies, also vs. the probit one, see Altman and Sabato (2007). A number of papers, among which Lennox (1999) and Altman and Sabato (2007), show that probit and logit models outperform DA model in default prediction.

To guarantee a reasonable level of precision, we use the number of eight-digit IPC classification codes reported in the patent document.

We follow the technology grouping proposed by the Observatoire des sciences et des techniques (OST). In particular there are 30 categories: 1 electrical devices - electrical engineering; 2 audiovisual technology; 3 telecommunications; 4 information technology; 5 semiconductors; 6 optics; 7 analysis, measurement, control; 8 medical engineering; 9 nuclear engineering; 10 organic fine chemicals; 11 macromolecular chemistry, polymers; 12 basic chemical processing, petrol; 13 surfaces, coatings; 14 materials, metallurgy; 15 biotechnology; 16 pharmaceuticals, cosmetics; 17 agriculture, food; 18 general processes; 19 handling, printing; 20 material processing; 21 agriculture & food machinery; 22 environment, pollution; 23 mechanical tools; 24 engines, pumps, turbines; 25 thermal techniques; 26 mechanical elements; 27 transport; 28 space technology, weapons; 29 consumer goods & equipment; and 30 civil engineering, building, mining.

See Stein (2007) for the selection of the out-of-sample data set.

Although the patent-value variable per se reduces the PD, for the sake of completeness we also try the interaction between this innovation variable and equity. As expected, the interaction term (not reported in the table) is not significant thus confirming that the patent’s value reduces firm riskiness independently of the equity level.

To be noted that the variance inflation factor is low for each of the explanatory variables: this level indicates that there are no problems with multicollinearity.

Overall, the size of the balance-sheet variables’ coefficients is stable and robust across all specifications.