Which Variables Determine the Accuracy of Default Probabilities Implied by Structural Models?

We argue that the Merton (1974) model's relatively high ability to forecast bankruptcy stems from its ability to capture either the chance of net worth dropping below an externally-imposed threshold or of an economic insolvency. Using unique bankruptcy data from fifteen countries, our evidence suggests that model-implied default risk estimates are more informative if a firm's net worth is constrained by covenants. In contrast, we only find weak evidence that model assumptions presumed to be important for capturing economic insolvency risk matter. Finally, asset liquidity and the efficiency of the variables used in calibration also relate to forecasting power