An empirical analysis of the KMVMerton model - A case of Swedish real estate companies
MetadataShow full item record
This thesis examines the contribution of applying the KMV-Merton model on Swedish real estate companies listed at the NASDAQ OMX Nordic Real Estate Index. Comparing the KMV-Merton model credit rating to frequently applied credit metrics, we find that the model adequately captures relevant information contained in these metrics. Additionally, the model proves robust when using long time series. Applying data from the time interval 2007-2014, we estimate econometric models to decompose significant predictor variables for credit spread variation at issuance. We obtain data directly from financial statements to assure statistically useful estimates. A univariate econometric model including the KMV-Merton default probability explains pooled cross-sectional regularities in credit spreads rather well. Combining firm financials, macroeconomic predictors and bond characteristics with the pure structural model, we conclude that a comprehensive hybrid model has improved fit. This result suggests that the KMV-Merton model is unable to capture all information contained in financial- and macroeconomic data. In particular, a model including the default probability, loan-to-value, the 3-month annualized interbank rate, coupon structure and credit rating is able to explain 80.19% of credit spread variation. Including a time variable enables us to exclude the existence of spurious time correlations and construct a model that is unconstrained in the parameters. Overall, the explanatory power achieved aligns with empirical research. In summary, we conclude that the KMV-Merton model yields significant statistics for credit risk assessment of Swedish real estate bonds at issuance. However, the statistic does not prove sufficient, as the comprehensive hybrid outperforms the univariate model.