1. Liquidity Horizon model was developed to predict the
liquidity of the traded credit products. 2. This was one model which enabled bank to comply with BASEL guidelines for IRC (Incremental Risk Charge) calculation. 3. Model addressed following tasks: • Examine the liquidity of traded credit products. • Consider the hedging impact on the liquidity period. • Statistical analysis of the market liquidity for the traded credit products. z Model Development steps:
Step 1: Defining the Objective
Step 2: Gathering the Data Step 3: Preparing the Data for Modeling Step 4: Selecting the Variables Step 5: Processing and Evaluating the Model Step 6: Validating the Model Step 7: Implementing and Maintaining the Model z Step 1: Defining the Objective
• The objective of LH model was to estimate liquidity
period for traded credit products (primarily Bonds, CDS & Loans)
• Comparing the actual Realized Time to Liquidation
(RTL) against the model generated liquidity period.
• Analyzing the hedging impact on the liquidity period of
the product. z Step 2: Gathering the Data
• The dataset consists of all the trades between the
period Oct 2008 till Sep 2009 and sourced from various trade systems.
• To get large volume of data, both live and dead trades
have been included.
• Data excluded all the trades where Credit Suisse was
the counterparty as this could potentially skew the results. z Step 3: Preparing the Data for Modeling
• Data Accuracy (unusual values, outliers).
• Data completeness (Are missing values important).
• Data imputation (for missing values)
• Data sampling (to select 30 random samples from each
traded credit products). z Step 4: Selecting the Variables
• Stepwise regression was performed to select the
significant variables for the model.
• Data was segmented in three product categories Bond,
CDS and Loan. z Step 5: Processing and Evaluating the Model • Linear regression model was used to estimate liquidity period for the traded credit products. • Bid/Ask spread was used as a proxy for market liquidity for the products. This was our response or dependent variable. • Bid/Ask spread was bucketed into 10D, 20D, 40D, 60D & 120D. • Primary response or predictor variables were Ratings, Maturity, Sector, Region and Currency. z Step 6: Validating the Model
• Actual realized time to liquidation is compared with
model predicted liquidity period for test sample.
• Residual/error analysis was done to examine the model
performance.
• Input sensitivity test: applying shift to independent
variables and compare the predicted liquidity with the baseline. z Step 7: Implementing and Maintaining the Model • Once model was developed and validated, it was implement within in the Front Office trading system . • User Acceptance Test(UAT) was performed to ensure model is implemented correctly and outcomes are sync with desired outcomes. • Entire process of model development and methodology was documented in model document. • Team was designated to monitor performance of the model on a periodic basis. Monitor performance of the models on a periodic basis z