Summary: Fitch's Equity Implied Ratings and Probability of Default (EIR) model is estimated to provide a view of a firm's credit condition given its current equity price and available financial information.
The Fitch EIR model is a proprietary hybrid probability of default and rating model that incorporates an option-based barrier model with hybrid adjustment of firm's financial information and market information. Our barrier-option based PD provides a forward-looking structural default probability. Changes in this structural default probability provide leading information about changes in the credit quality of a debt issuer, and thus help to understand impending rating change and default. The model makes use of a small, but very carefully selected subset of accounting and market variables.
Fitch's proprietary default database contains over 7,900 default globally, spanning from the 1960s through 2006. Model calibration for Fitch's EIR model is based on data from 1990 through 2005 for North American firms, and for the period between 2000 through 2005 for global firms using daily price history and annual financial statements.
The Fitch EIR model covers approximately 27,000 entities globally: 13,000 in the US and Canada, plus another 14,000 firms from more than 70 other countries. For all firms, the model provides daily output of estimated default probability (PD) for both one-year and five-year horizons, and the implied agency rating (IR).
Accuracy Ratio (AR) power of the model for the rated universe is 86.4% and 68.8% for 1-year and 5-year horizons, respectively. When compared to established benchmark models, the EIR model outperforms peer models by wide margins. Furthermore, it is shown that our hybrid methodology (whereby certain financial ratios and market information are combined with a pure option-based barrier model) outperforms the pure option-based barrier model's AR by 4% and 6% at the 1- and 5-year horizons, respectively.
In addition to AR, a battery of tests are performed to measure the performance of the model, including k-fold tests, walk-forward tests, and lead-lag analysis. The results of these tests reveal that the model is very powerful in distinguishing good credits from bad credits, maintains robust performance through time and is applicable across different industries, geographic regions, and firms with different sizes.
