MODELING METHODOLOGY DECEMBER 2015DECEMBER 2015CAPITAL MARKETS RESEARCHStressed EDFTM Model Development, Implementation,&ValidationSummary Stressed EDFTM (Expected Default Frequency) credit measures are conditional, one-year default probability forecasts for public companies given a hypothetical economic scenario. They are used primarily by financial institutions to stress test their C&I loan, corporate bond, and CLO portfolios and by portfolio managers to understand risk using scenario analysis. Derived from Moodys Analytics Public Firm EDF measures, Stressed EDF measures can substitute seamlessly fortraditional point-in-time PDs whenever it is necessary to assess credit risk in alternative, future macroeconomic conditions. The Stressed EDF methodology relies on estimates of structural relationships between key macroeconomic risk factors and public firm EDF measures observed over a long history. Users seeking to understand only the models basic underpinnings and applications should refer to Stressed EDF Measures At a Glance. Here, we provide a more rigorous description of the development, implementation, and validation of the Stressed EDF9 model. In accordance with industry standards of model risk management, we document controls for data and code integrity, modeling considerations and decisions made during development, andthe results of sensitivity and validation analyses in addition to a description of the final model mechanics.Authors Danielle H. Ferry, PhD +1.212.553.7781 danielle.ferry@moodys.comGlenn Levine +1.212.553.9595 glenn.levine@moodys.comContact Us Americas +1.212.553.1658 clientservices@moodys.comEurope +44.20.7772.5454 clientservices.emea@moodys.comAsia (Excluding Japan) +852.3551.3077 clientservices.asia@moodys.comJapan +81.3.5408.4100 clientservices.japan@moodys.com2 DECEMBER 2015CAPITAL MARKETS RESEARCHTable of Contents1 Introduction 32 Data and Methods: Estimation 32.1 Economic Data 32.2 Firm-Level Data 42.3 Input Data Processes 52.3.1 Data Discontinuities 62.3.2 Missing Data 62.4 Model Design 62.4.1 Modeling Object: Distance-to-Default 72.4.2 Macroeconomic Risk Factors 72.4.3 Top-Down Model 92.4.4 Bottoms-Up Model 123 Data and Methods: Projections 133.1 Off-the-Shelf Economic Scenarios 133.2 Custom Economic Scenarios 163.2.1 Moodys Multi-Country Economic Scenario Accelerator (MCMESA) 163.2.2 User Inputs&Data Processing 183.2.3 Methodological Differences From Off-the-shelf Stressed EDF Measures 194 Stressed EDF Model Testing and Validation 194.1 Model Testing 194.2 Perfect Foresight Exercise 194.2.1 Out-of-Sample 204.2.2 In-Sample 224.3 Sensitivity Analyses 244.3.1 Macroeconomic Sensitivity 244.3.2 Autoregressive Sensitivity 254.4 Model Updates 265 Conclusion 27References 283 DECEMBER 2015CAPITAL MARKETS RESEARCH1 Introduction Stressed EDFTM (Expected Default Frequ...
