Loan Default Analysis: A Case for CECL Tuesday, June 12, :30 pm

Transcription

1 Loan Default Analysis: A Case for CECL Tuesday, June 12, :30 pm Insert Your Photo Here If no photo is available, center contact details on page. Presented by: Guo Chen Director, Quantitative Research ZM Financial Systems 1020 Southhill Drive, Ste. 200 Cary, NC P: E: Guo.Chen@zmfs.com

2 Overview More requirements around capital adequacy analysis from various regulatory bodies Data review and preparation - Multiple data sources - Data cleaning and standardization - Loan segmentation Loan analytics and modeling Historical performance by multiple dimensions (e.g., loan type, vintage) - Migration matrix - Regression models

3 The Data A multi-billion-dollar bank s loan portfolio monthly loan performance history - Data between early 2002 and May about 40k+ individual loans - over 1.3 million observations The whole data set covers various loan profiles, including: - Commercial loans (~40%); - Real estate loans (~15%); - Construction loans (~12%); - Consumer loans (~8%); and - Home equity lines (~6%) Public economic data such as HPA, Unemployment, and GDP etc. v

4 Data Challenges Only performance data available. No access to common loan credit indicators, such as debt service coverage ratio (DSCR), loan-to-value (LTV), etc. Missing data for lots of loans. For example, about 30% of the loans do not have their STATE field populated Some data provides conflicting information Certain complicated situations, e.g. loan restructuring, cannot really be captured by just checking the loan performance records. Data cleaning is a interactive process, working closely with the bank

5 Loan Data Scrubbing Accounting standards require disclosure of loan segments. The bank already divided its loan portfolio into several segments: - Consumer loans - Commercial loans - Real Estate loans - Construction - Lease Disaggregate segments into classes with similar risk characteristics, generally represent different loan types Further disaggregate classes into pools of loans where appropriate (ASC )

6 Loan Performance by Multiple Dimensions

7

8

9

10 Historical Loan Default Vectors Loan loss rates by different dimensions give a good picture of historical loan performance. With good history data, we can do more and derive the historical default vectors. These vectors can be easily applied to a discounted cash flow method (DCF) in loss analysis:

11

12

13

14 Migration Matrix Delinquency status are updated monthly in the performance data set Together with the loan default status, we estimate the loan status migration probabilities for Commercial Term loans Current Dlq30 Dlq60 Dlq90 Dlq120 Dlq150 Dlq180 Default Current Dlq Dlq Dlq Dlq Dlq Dlq

15 Migration Matrix Multi-period default probabilities can be computed by means of matrix multiplication. Assuming we have a loan that s 60days delinquent Initial Status Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Current % 67.92% 81.07% 88.65% 94.07% 96.75% 97.93% 98.42% Dlq % 5.21% 2.88% 1.98% 1.39% 1.10% 0.97% 0.91% Dlq60 100% 10.28% 5.64% 2.31% 1.06% 0.57% 0.34% 0.24% 0.19% Dlq % 9.29% 3.92% 1.64% 0.72% 0.34% 0.19% 0.12% Dlq % 6.89% 3.34% 1.46% 0.62% 0.27% 0.12% 0.06% Dlq % 0.50% 3.11% 1.49% 0.65% 0.27% 0.12% 0.05% Dlq % 0.52% 0.36% 1.25% 0.72% 0.33% 0.14% 0.06% Default % 4.03% 3.02% 2.47% 1.28% 0.61% 0.30% 0.17%

16 Migration Matrix Setting up the migration matrix in ZMdesk

17 Regression Models Logistic regression is a popular choice among practitioners Can incorporate various economic scenarios into the forecast Easy to implement: once the parameters have been estimated, the regression result be evaluated in a straightforward way Starting the model construction with fitting a full model including all potential predictors Only those can sufficiently discriminate between default and non-default are included in the final model.

18 Regression Models Applied regression analysis to the Commercial R/E term loans The bank-assigned internal loan rating (WCHCOD) is an obvious candidate of the model input The three best internal ratings 1, 2, and 3 don t discriminate each other in the model, so they are put into one rating group A in the model HPA rate is the macroeconomic input to the final model. It s derived from the Seasonally-Adjusted Purchase-Only Index from the FHFA website The 15 month lagged HPA gives the best model fitting result, it also has the highest correlation with the loan default events

19

20 Regression Models A random sample of about 70% of the Commercial R/E Term Loan data are used for model training to estimate the model parameters The remaining 30% out-of-sample loans are used for model validation Only HPA and the internal rating are kept in the final model. They are more stable and retains the similar predictive power of the full model. PD = 1+e P 0+P1 HP A Lag 18 +P2 Rating 1 We compare results of in-sample and out-of-sample fitting

21

22

23 Regression Models Assuming a loan issued in April 2013, with rating A, we can manually calculate the model projected default rates Loan Specs Model Parameters Calculation IssueDate Intercept Period HPA_lag15 PD(MDR) CDR Rating A HPA_lag % 0.52% Rating % 0.67% % 0.53% % 0.43% % 0.31% % 0.31% % 0.29%

24 Regression Models It is straight forward to set up the model in ZMdesk

25 Regression Models Assign the model to a loan

26

27 Regression Models Based on the related information in the Default result tab, we can get the following CECL measures assuming a loan for $1,000, CDR first 12 months 0.34% - CDR Life 0.2% - Total dollars in default -- $8,700 - Total dollars recovered -- $1,000 - Net loss in dollars (undiscounted) -- $7,700 - Net loss (discounted basis at book yield) -- $7,100

28 Regression Models The simulated loan cash flow:

29 Regression Models Logistic regression model explains the default event within a certain period Cox Proportional Hazards (CPH) models link the time-to-default under consideration of censoring (meaning observation stopped due to maturity, prepayment, or default) Investigated the Commercial Term Loan portfolio with the CPH model. The model variables are - The internal loan rating - 12-month lagged HPA - Loan amount as categorical input: Small: Large

30 Regression Models The simulated loan cash flow: Parameter Estimate Std. Error Chi-Square Pr > ChiSq Hazard Ratio Rating B Rating C < Rating D < Rating M < ORGAMT Small < HPA_lag <

31 Regression Models The hazard ratio of Rating B says loans with the rating B have 2.2% more chance to default when compared with loans with rating A. Loans with ORGAMT in the Small category is times more likely to default. A 1% increase in the HPA rate implies a 10.9% decrease in PD Yet to be integrated into ZMdesk

32 Q&A