MODELLING CREDIT RISK FOR INNOVATIVE FIRMS: THE ROLE OF INNOVATION MEASURES

Transcription

1 MODELLING CREDIT RISK FOR INNOVATIVE FIRMS: THE ROLE OF INNOVATION MEASURES Grid Thoma (University of Camerino and CEFIN) with Costanza Torricelli (University of Modena and Reggio Emilia, CEFIN) and Chiara Pederzoli (University of Milano-Bicocca and CEFIN)

2 Plan of the presentation Motivation and goal The Dataset The Model Accounting variables selection Definition and measurement of innovation variables Results: the contribution of innovation variables to PD estimation Conclusions 2

3 Motivation Disruptive role of innovative SME in the Knowledge Economy Financial constrains for SMEs: particular severe for R&D projects (Hall & Lerner,200;EIB,2009): Financial distress in an early stage In EU, the VC industry is underdeveloped Credit quality is harder to gauge 3

4 Motimation (cont) Basel II IRB: rating assignment and quantification, PD as an input for the capital function In Basel II separate treatment of exposures to SME (Small- and Medium-sized enterprises) borrowers Firm-size adjustment allowed in minimum capital requirement under Pillar I Definition of SME exposures: sales for the consolidated group of which the firm is a part less than 50 milion euro 4

5 Goal of the paper Contribute to credit risk assessment of innovative SMEs by considering the signaling value of some innovative assets (Harhoff et al. 2009, Hsu and Ziedonis, 2009) Use one of the most used model for the PD of SMEs, i.e. the logit with accounting variables Evaluate the improvement for the PD estimation obtained by adding knowledgebased variables Estimate on a unique dataset of innovative firms 5

6 Dataset Def. innovative firm: at least one patent application in EPO/PCT systems (PATSTAT database & EPO Bulletin) SME selection: -50 mln sales (EC definition); EU5 + CH + Nor Overall sectors, excluding utilities, transportation, and finance Matching with Amadeus business directory (Thoma et al. 200) Demographic & accounting information, such as sector, ownership, balance sheet, profit & loss account, firm s IPTS legal workshop, status June 20 6

7 Bankrupt firms Def. of default and data availability Amadeus: active and bankrupt (a) unable to pay the creditors; (b) the assets are held by a receiver; (c) assets and property of the company redistributed. 7

8 The Literature on default prediction based on financial ratios Seminal papers: Beaver (966), Altman (968) (univariate and multivariate DA) Logit model: Ohlson (980) Beaver (2005): balance-sheet financial ratios still preserve their predictive ability; market-based variables partly encompass accounting data Recent works using logit for SMEs: Altman and Sabato 2007 (Australia), Behr and Güttler 2007 (Germany), Figni and Fantazzini 2009 (Germany), Fidrmuc and Heinz (Slovakia) and Pederzoli and Torricelli 200 (Italy) 8

9 The Logit model X X PD Y P R k t ik k R k t ik k i t i = = = = = +,,, ) exp( ) exp( ) ( β α β α in t i obligor for regressor k X in t defaults not does i obligor if in t defaults i obligor if Y th n i t ik n i t i = + + = = = +,...,,,...,, 0 9

10 Variables: financial ratios (Altman & Sabato, 2007) LEVERAGE Current Debt/Equity Equity/Debt Debt/Assets LIQUIDITY Cash/Assets Cash/Sales Stocks+Tradedebt/Assets COVERAGE EBITDA/Interest EBIT/Interest PROFITABILITY EBIT/Sales EBITDA/Assets Profits/Assets Earnings/Assets Profits/Sales ACTIVITY Sales/Debt Tradedebt/Sales Tradedebt/Debt 0

11 Knowledge variables R&D intensity Patents/R&D personnel Patent value built with a maximum likelihood factor index with three indicators: Potential market size designated countries Technological relevance citations received Technological scope # of patent classes Time dimension of knowledge: capitalize variables with perpetual inventory method K t = (-δ) K t- + I t ; δ = 5% depreciation rate

12 Censoring issues Publication lag (8 months) Priority lag (2 months) Citation lag (3 years citation window) Technology fixed effects (30 macro areas) INPADOC family patents 2

13 Estimation Pooled cross section estimator of the PD Default years: =>Fiscal years: Sample: random selection up to 6% of defaulted firms Logit regression model Selection with backward stepwise regression In-sample vs Out-sample 3

14 Variables: financial ratios LEVERAGE Current Debt/Equity Equity/Debt Debt/Assets LIQUIDITY Cash/Assets Cash/Sales Stocks+Tradedebt/Assets COVERAGE EBITDA/Interest EBIT/Interest PROFITABILITY EBIT/Sales EBITDA/Assets Profits/Assets Earnings/Assets Profits/Sales ACTIVITY Sales/Debt Tradedebt/Sales Tradedebt/Debt Sales 4

15 5

16 Results in-sample (776 obs) Notes: Each regression includes fixed effects at the level country, sector, cohort, year and listed/not listed. 6

17 Model evaluation: the CAP curve BCBS (2005) Studies on the validation of internal rating systems: Percentage of defaulted perfect model tested model random model DR Percentage of total Accuracy ratio (AR) = area under the tested model curve / area under the perfect model curve 7

18 8

19 +2.% +5.% 9

20 Error Type I & II in-sample vs out-sample 60% 50% In sample Out sample 40% Erro or Type I 30% 20% 0% 0% 7% 32% 65% 84% 89% 92% 95% 96% 96% 99% Error Type II 20

21 Conclusions We confirm the validity of limited dependent variable model with financial ratios to represent default event as for standard SMEs Standard accounting variables: role confirmed with expected signs Patenting and risk assessment Patent-related variables improve prediction: Patent value decreases PD Patent stock per se is not a value Patent stock decreases PD only if equity is appropriate 2