DFAST STRESS-TESTING

The Quantal Stress Test Engine (STE) uses the scenarios specified by US banking regulators for stress testing by US banks under the Dodd Frank Act Stress Test and the Comprehensive Capital Analysis and Review (DFAST-CCAR).

 

The STE answers the following questions:  Suppose that the state of the economy is expressed in terms of one of the three DFAST-CCAR scenarios (Supervisory Adverse, Supervisory Severely Adverse, and Supervisory Baseline), or a user scenario that specifies a future path of quarterly values for the 28 DFAST-CCAR macroeconomic variables.   What is the impact on the likelihood of default for each loan within each quarter, and what is the likely loss given default (LGD) consequence for each loan (or class of personal loans) given the projected macroeconomic environment? The Stress Test Engine assesses this impact relative to the user’s baseline current estimated default probabilities across all possible future scenarios as of the scenario start date.

 

We calculate the impact of each stress scenario on the PDs and LGDs of the bank’s existing loan portfolio (we do not model how a bank might change the composition of its loans during a period of financial stress, though a client plan could be incorporated). The figure below shows a “bottom up” term structure of stressed PDs, as of the end of September 30, 2012, for the adverse and severely adverse scenarios for a sample of 3,338 publicly traded U.S. companies. Here, the bank input PD is that published for September 30, 2012 by the National University of Singapore’s (NUS) Risk Management Institute (RMI) for each company. The baseline, adverse, and severely adverse scenarios are those given in the November 15, 2012 CCAR report. The User1 scenario, which we’ve labeled a “Depression” scenario, takes the CCAR Severely Adverse Scenario and eliminates the recovery phase, i.e. it has flat real GDP growth after the initial “bad” shocks, along with deflation, low interest rates and a declining stock market.

Median PDs in the alternative scenarios for 1 to 5 years out for a sample of 3,338 publicly-traded companies for which PDs are available in the NUS RMI database

Example of Projected Quarter-by-Quarter Changes for CCAR Scenarios over Period 2012 Q4 to 2015 Q4

For personal loans and credit card debt, we then predict that the quarter-to-quarter change in charge-off rates in a Bank rating category is related to the respective change in the Federal Reserve rate via the following error correction model (ECM) specification:

 

 

In this ECM specification, the change for a loan category will be higher if the charge-off rate for a given category in quarter t is “low” relative to what it would normally be given the Federal Reserve national rate. For category 7 Prosper Loans:

A  “tops-down” could also be applied in stress-testing. For example, the following figure shows the model’s forecast for the quarter-by-quarter asset (“EBITDA cash flow”) values of companies in the Basic Materials industry, starting in Q4 2012 and ending in Q4 2015, given the following scenarios: (1) the CCAR baseline (Green); (2) the CCAR adverse (Orange); (3) the CCAR severely adverse (Red); and (4) an arbitrary “Depression”-User-specified scenario:

Copyright © Quantal International, Inc.

Quantal International, Inc.
455 Market Street
Suite 1200
San Francisco, CA 94105
Phone: 415-644-0754 Fax: 415-644-0948