How We Help
CDG can assist you in preparing or validating your loss forecasts. We also can drill down into your portfolio to understand risk and performance by segment. Combined with account profitability data high risk and low reward parts of your portfolios can be identified and addressed.
As the name implies loss forecasting is the process of estimating future credit losses. There are 3 primary approaches to consider to loss forecasting today which can support the new CECL requirements that will be effective beginning in 2020.
Roll Rate Forecast
A roll rate forecast estimates the balances that “roll” from one delinquency level to the next each month. The roll rate percentages are based on historical experience and take into account factors such as the number of days in the month, seasonality, and any significant balance growth.
- Simple to calculate
- Most accurate in short-term
- Limited data required (essentially balances by delinquency bucket)
- Relies on balance forecasts
- Unreliable 9+ months out
This method is most accurate for stable, mature portfolios. However, its accuracy diminishes over time and as new portfolio segments are added into the mix.
A vintage forecast tracks performance of discrete time slices (based on origination date) of accounts and projects future behavior by using the actual performance of older accounts to estimate losses on newer accounts.
- Most accurate in portfolios with significant chunks or vintages of new accounts
- Provides detailed forecasts by portfolio segment
- Compares performance of different segments and vintages to highlight potential problems and opportunities in your business
- Relies on significant historical data
- Can be time-consuming to prepare and update
- Requires thorough documentation of portfolio changes and assumptions
Vintage forecasts represent a significant jump in analytics from roll rate forecasts. The insights you can glean from the detailed analysis can lead to improvements ranging from small steps to giant leaps.
Risk Level Forecasting
This approach estimates future losses by using the probability of default (PD) and loss given default (LGD). Calculated at the account level, both PD and LGD are based on statistical modeling and take into account the age of a given account. The accuracy of these forecasts are based on the accuracy of these models for the current set of accounts in the portfolio. For example, if the accounts used to develop the models mirror the current porfolio the accuracy will be high. If the current profile differs from the historical one variations are bound to occur.
- Can be highly accurate with a homogeneous portfolio
- Account level estimates can be deployed in other strategies
- Relies on significant historical data
- Requires statistical modeling skills with on-going model validation
This approach is generally restricted to the most sophisticated (or analytically ambitious) lenders.
Current Estimate of Credit Losses (CECL)
CECL is the newly issued (June 2016) accounting standard that requires that at each reporting period all future losses from existing accounts must be included in the allowance for credit losses. This represents a significant change in how losses are accounted for and will likely require a significant P&L hit when first implemented. However, all financial institutions will see similar impacts.
Unfortunately detailed requirements have not been established. It will require a significant change in how data is collected and managed in order to support CECL loss forecasts for most lenders. The best way to think about this is as a long-terms vintage forecast that needs to take into account segment details (by account type and risk level), the age of accounts, and balance persistency. This can be expanded to include probability of default and loss given default.
Getting ready for CECL will require significant planning and preparation to demonstrate the accuracy of the forecasting methodology. No fun but not time like the present to get started.