Modelling Probability of default of companies is of the utmost importance in the area of risk management. In this thesis, we describe a method for estimating the probability of default via the usage of vine copulas and the companies' capital simulation, as presented in Valle_2016. The stand out feature of this method is that it uses only balance sheet data, as it needs only short and long term assets and liabilities. Additionally, some of the Merton model assumptions can be discarded, which enables its usage for small and medium enterprises, which comprise a significant part of Slovenia's economy.
As forecasting the probability of default under certain conditions is sometimes as important as estimating it in the present, we expanded the model with conditional vine copulas. This expansion enables forecasting with additional parameters, which can be shocked for the usage in stress testing. We used both methods on ten years of real data for the company Petrol d.d. For the expanded method, we used the yearly change of GDP for Slovenia as an additional variable. It turns out that it works as expected, as the simulated own funds were smaller in the case of lowered GDP. Additionally, we have shown that the performance of the method improves, and the method becomes theoretically sounder if we use relative changes of real data over some timeframe instead of absolute valued data.
|
