This thesis aims to examine the dependence of bank loan recovery rates on various factors, on the entity level. Especially after the last financial crisis, the modeling of recovery rates is considered to be a vital topic for banks. Having a good model, the banks are able to predict, already on the default date, the value of the recovery rate at the end of the resolution time. This study is based on the largest bank loan credit default dataset, provided by Global Credit Data. With a main focus on small and medium enterprises located in the United States, values of recovery rates outside the open interval (0; 1) are examined separately from the rest of the cases. The set of recovery rates in (0; 1) is split based on a binary crisis indicator, which we model using monthly macroeconomic data. The crisis indicator has predictive power, indicating whether turbulent periods are expected to occur within the resolution time or not. Combining the different parts of the analysis, we finally suggest two tree models for small and medium enterprises in the United States, which include logistic regression, linear regression and/or quantile regression. One main conclusion of this thesis is that splitting the recovery rates reported in (0; 1) into two subsets based on the crisis indicator, reveals different relation between the recovery rate and the various factors in each subset. Another important observation is that the tree models are able to predict recovery rates more accurately comparing to the models built upon recovery rates reported in (0; 1).     «

This thesis aims to examine the dependence of bank loan recovery rates on various factors, on the entity level. Especially after the last financial crisis, the modeling of recovery rates is considered to be a vital topic for banks. Having a good model, the banks are able to predict, already on the default date, the value of the recovery rate at the end of the resolution time. This study is based on the largest bank loan credit default dataset, provided by Global Credit Data. With a main focus on...     »