Whats New with RiskCalc Plus and Private EDFs? Douglas Dwyer, - - PowerPoint PPT Presentation
Whats New with RiskCalc Plus and Private EDFs? Douglas Dwyer, Managing Director, Single Obligor Research Janet Zhao, Associate Director, Single Obligor Research Mehna Raissi , Associate Director, Single Obligor Product Management 2 Agenda 1.
Douglas Dwyer, Managing Director, Single Obligor Research Janet Zhao, Associate Director, Single Obligor Research Mehna Raissi, Associate Director, Single Obligor Product Management
What’s New in RiskCalc Plus and private EDFs? October 2011
2
What’s New in RiskCalc Plus and private EDFs? October 2011
What’s New in RiskCalc Plus and private EDFs? October 2011 4
» A third generation econometric model for private firm default risk using your company-specific financial statements
» Extensive explanatory analytics » A well understood, accepted standard » Market driven, industry level credit cycle adjustment » The largest, cleanest private firm default database in the world
What’s New in RiskCalc Plus and private EDFs? October 2011
» All RiskCalc models are created using Moody’s Analytics Credit Research Database (CRD) » The CRD is a collection of credit risk data consisting of financial statements, descriptive customer information, and defaults events for 28 countries, covering most of the world’s GDP » The CRD collects data from institutions (banks and non-banks) and vended sources, then runs the data through rigorous analysis and data cleansing steps » The CRD also provides participating institutions with data cleansing and data quality reporting, benchmark reporting, and RiskCalc scoring services
What’s New in RiskCalc Plus and private EDFs? October 2011
What’s New in RiskCalc Plus and private EDFs? October 2011
7
What’s New in RiskCalc Plus and private EDFs? October 2011 8
What’s New in RiskCalc Plus and private EDFs? October 2011
What’s New in RiskCalc Plus and private EDFs? October 2011
10
Quality of Management
What’s New in RiskCalc Plus and private EDFs? October 2011
11
What’s New in RiskCalc Plus and private EDFs? October 2011
12
Example firm’s PD level climbed up before default PD levels of the peer group The Peer Group: Firms in Trade Sector
What’s New in RiskCalc Plus and private EDFs? October 2011
13
What’s New in RiskCalc Plus and private EDFs? October 2011
What’s New in RiskCalc Plus and private EDFs? October 2011
15
What’s New in RiskCalc Plus and private EDFs? October 2011
» We look at the following aspects when validating a model
– Is the model working as intended? – Is the discriminatory power being maintained? – Is the level of the PD appropriate? – Can the model be improved?
» The recent validation combines financial statement information with loan accounting system information to track cohorts over time
– Data provided by 11 financial institutions – Data runs from 1999 through 2010
» The RiskCalc U.S. model performs well based on the recent data
16
RC US v3.1 Z-Score Defaults Firms Number
53.6% 37.7% 5,134 81,601 260,959
What’s New in RiskCalc Plus and private EDFs? October 2011
» Extend the model to cover: non-for-profit organizations, real estate operators, and potentially dealership firms – Unique accounting standards, especially for non-for-profit organizations – Key risk drivers behave differently from those of a typical corporate – Central default tendency differs from that of the US corporate model » Non-for-profit model uses inputs specific for this asset class » Models built specifically for the three asset classes improve the accuracy ratio » Non-for-profit firms have lower default rate, while real estate operators and dealership firms have higher default rate
17
New Model US 3.1 Non For Profit 59% 53% Real Estate 56% 44% Dealership 58% 48%
New Models Improve Accuracy Ratio
What’s New in RiskCalc Plus and private EDFs? October 2011
» Stress Testing » Credit line usage and exposure at default » Behavioral Overlay » Accounting Quality
18
What’s New in RiskCalc Plus and private EDFs? October 2011
19
SalesGrowth =F(GDP, interest rate, …) Estimate Other RiskCalc Inputs based on the Sales Estimation Score Estimated Inputs Through RiskCalc
Next Period Input Computation Sales(t+1) Sales(t)* (1+ b0 + b1*X(t)), X(t) are macroeconomic variables Net Income(t+1) Net Income(t) + g1 *(Sales(t+1)-Sales(t)) Retained Earnings(t+1) Retained Earnings(t) +Income(t+1) Total Assets(t+1) Total Assets(t)+Income(t+1) Cash(t+1) Max(0,Cash(t)+Income(t+1)) EBITDA(t+1) EBITDA(t)+ ∆Net Income
Moody’s Economy.com
What’s New in RiskCalc Plus and private EDFs? October 2011
» We pool credit line usage data together from eight financial institutions in the U.S. » The sample covers approximately 7,600 defaulters and 134,000 non-defaulters with valid usage information, 2000–2010 » The credit line usage of defaulted firms are indeed higher than that of non-defaulted firms
20
40% 50% 60% 70% 80% 90% 100%
Usage Ratio Quarter Prior to Default mean - Defaulters median - Defaulters mean - NonDefaulters median - NonDefaulters
What’s New in RiskCalc Plus and private EDFs? October 2011
21
» Usage increases with default probability measured by RiskCalc EDF » Evidence shows that banks do monitor credit lines. Usage relates to collateral types, bank internal rating and commitment size » Both commitment and balance amount declined during the recession times
0% 2% 4% 6% Sep-01 Mar-02 Sep-02 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 Sep-06 Mar-07 Sep-07 Mar-08 Sep-08 Mar-09 Sep-09 Mar-10 Change in Commitment or Balance Commitment Change Balance Change
What’s New in RiskCalc Plus and private EDFs? October 2011
» RiskCalc is based on financial statement fields » Behavior factors may also add to default prediction: credit line drawdown, age of lending relationship, freshness of financial statement… » We can potentially add a behavioral layer to RiskCalc models » Credit line drawdown add to the power of default prediction; more so for small firms, firms with Pass internal rating, firms with stale financial statements
Variable AR EDF only 53% Usage only 39% 70% weight on EDF, 30% on Usage 57%
Combining Credit Line Usage with EDF Improves Accuracy Ratio
What’s New in RiskCalc Plus and private EDFs? October 2011
» RiskCalc is based on financial statement items. Financial quality metrics can potentially help RiskCalc users to identify problematic financial statement for further review » For public firms, researchers validate financial quality measurement with restatement event, SEC enforcement, or accounting-related lawsuit » For private firms, how can we validate the accounting quality measurements? » We construct a set of variables, and examine whether the identified “abnormal” statements are
– Less useful in predicting defaults – Less predictive of future cash flows and earnings – Less likely to be audited – Less likely to be associated with loan accounting records
23
What’s New in RiskCalc Plus and private EDFs? October 2011
» The set of variables include accruals measurements, unexplained retained earnings, together with variables related with operating margin, sales to assets, depreciation to plant property and equipment (PPE)
– Accrual is the non-cash component of income, it is more subjective than the cash component of income – For each industry and year, we identify top 10% firms with high accruals
» The “abnormal” statements have lower power in predicting defaults, are less predictive of next year’s earnings and cash flows, are less likely to be audited, and are less likely to be associated with loan accounting data
24
Group Accuracy Ratio normal 50% abnormal 32% Matched normal 49% Group Corr (ROA0, ROA1) Corr (ROA0, CFO1) Normal 0.64 0.54 Abnormal 0.57 0.49
What’s New in RiskCalc Plus and private EDFs? October 2011
What’s New in RiskCalc Plus and private EDFs? October 2011
1) Does your organization have exposures to dealerships? A) Yes, we have exposures and currently use a proxy model that works fine B) Yes, we have exposures and currently use a proxy model. We are interested to learn more about RiskCalc for dealerships C) No, we do not have exposures to dealerships and have no need for this specific model 2) Which topic are interested to learn more about? A) Stress Testing based on pre-defined macro-economic drivers B) Asset Class Expansion C) Non-Financial & Behavioral Overlay D) New Models & Validations
26
What’s New in RiskCalc Plus and private EDFs? October 2011
3) What’s your primary use of RiskCalc? A) As an input to internal rating model B) Benchmark to internal rating model C) Early warning tool D) Calculating loss reserves/capital 4) What’s your secondary use of RiskCalc?
A) As an input to internal rating model
B) Benchmark to internal rating model C) Early warning tool D) Calculating loss reserves/capital
27
What’s New in RiskCalc Plus and private EDFs? October 2011
Your feedback is valuable to us! Join the RiskCalc User Group for updates on the latest product and research
Contact Mehna Raissi for more information mehna.raissi@moodys.com
What’s New in RiskCalc Plus and private EDFs? October 2011
moodys.com
Doug Dwyer Managing Director, Research Douglas.Dwyer@moodys.com Janet Zhao Associate Director, Research Janet.Zhao@moodys.com Mehna Raissi Associate Director, Product Management Mehna.Raissi@moodys.com Anuj Gupta Director, Product Management Anuj.Gupta@moodys.com