Risk management for »peer to peer« lending

Nove digitalne tehnologije botrujejo procesu preoblikovanja obstoječih vrednostih verig finančnih produktov oz. storitev. »P2P« posojila so nov in inovativen način tako investiranja presežkov finančnih sredstev kot tudi prejemanja finančnega kapitala. Število tovrstnih posojil konstantno raste, vendar posojilodajalci niso profesionalni investitorji. Posojilodajalci prevzemajo veliko tveganje, saj so »P2P« posojila izdana brez zavarovanja. V ta namen »P2P« platforme izdajajo historične podatke o posojilojemalcih. V delu se osredotočamo na identifikacijo tveganj, ki so prisotna pri tovrstnem investiranju in na napovedovanje možnosti neplačil posojil. Empirična študija analizira podatke pridobljene iz platforme Bondora (N=1823) od leta 2009 do 2015. Opravili smo statistično analizo spremenljivk. Razvili smo Logit model za napovedovanje neplačil. Kakovost modela smo preverjali z ROC krivuljo, optimizacijo modela pa na osnovi uravnoteženja klasifikacijske natančnosti, kjer smo dololčili optimalno presečno vrednost. Rezultati so pokazali, da kreditni model za napovedovanje neplačil zmanjšuje verjetnost finančne izgube pri »P2P« investiranju.New digital technology leads to the process of transformation of the existing value chains and financial products or services. "P2P" lending is a new and innovative way as an investing surplus of funds as well as receiving financial capital. The number of such loans constantly grows, but lenders are not professional investors. Lenders take a lot of risks, as "P2P" lending is issued without collateral. For this purpose, "P2P" platform issue of historical data on borrowers. Research focuses on the identification of risks that are present in such an investment and to predict the possibility of loan default. The empirical study analyzes data obtained from the platform Bondora (N = 1823) from 2009 to 2015. We carried out a statistical analysis of the variables. We have developed a Logit model to predict loan defaults. The quality of the model was measured by ROC curve optimization model on the basis of balancing the classification accuracy, where we determined optimal cut-off value. The results showed that the credit default prediction model reduces the probability of financial loss on the "P2P" investment