S&P Global Market Intelligence recently recalibrated PD Model Market Signals (PDMS) for both corporate companies and financial institutions to reflect the latest default experience and market behavior since the model's inception in 2011. At the same time, we managed to decrease the noise typically seen in volatile markets without compromising the level of accuracy observed during the 2016 validation, prior to the recalibration. This is particularly useful for investment managers seeking clearer market signals and for risk managers looking at reliable early warning signals of credit risk increase.
In the “Model Recalibration” section, we summarize the model framework and list in detail the features that differentiate the recalibrated version (v1.1) from the previous release's version (v.1.0). In the “Impact Analysis” section, we discuss the effect of the recalibration on the model outputs (PD level and volatility), the model discriminatory power, and the migration statistics.PD Model Market Signals builds on a structural framework, using the market capitalization and the total liabilities of a publicly-listed company as initial inputs to calculate the asset value, the asset volatility, and the firm’s leverage. These three items are used to derive an intermediate model output, the Distance to Default (DD), which is empirically mapped to a PD, then further adjusted by industry, country, and sovereign risks, as well as size effects.
PD Model Market Signals consists of two modules: one for Corporates and one for Financial Institutions. We have recalibrated both modules, adding 2012-2015 data to expand our reference sample across both time and cross-sectional, and to make it more relevant to the current economic conditions, since v1.0 was trained based on data through 2011.
At the same time, we ensure to achieve smoother PD volatilities and reduced impact of market return outliers. Compared to fundamentals-based PD models, equity-based models are more prone to volatile outputs, often due to the underlying stock-market dynamics; this often causes spurious early warning signals about credit risk increase. As discussed in the Technical Reference Guides, we used a number of techniques (such as the robust volatility treatment) to cure the model outputs “noise”, but recent feedback from clients has paved the way for further enhancements.
In practice, we achieve the recalibration in two steps:
- First, we revisit the empirical mapping between the Distance-to-Default (DD) and the Probability of Default (PD) values: intuitively, a less steep relationship between DD and PD will naturally decrease the volatility of the PD outputs from both modules, given the same fluctuations of the DD values.
- Second, we update the DD scaling by industry, to account for different observed default rate (ODR) by sector in each module, and update the adjustments applied to the PD of Financial institutions, with a further improvement of the PD volatility.