Historical vs. Loan Age vs. Deal Age Analysis Pages

dv01 provides multiple ways for users to evaluate common performance metrics across different normalized periods.

Historical Analysis

For Historical Analysis charts, data is normalized by calendar month. These views can be leveraged to identify trends in seasonality and broader market events that may impact performance behavior at specific points in time. The table below demonstrates how Historical Analysis aligns end-of-month snapshots across different origination vintages to calculate various performance metrics as of a specific point in time for a hypothetical securitization with a cut-off date of January 15th, 2019 and a first collection period end date of February 28th, 2019.

Deal Age Analysis

For Deal Age Analysis charts, data is normalized by months since the cut-off date for a given transaction. Similar to Historical Analysis charts, these views can be leveraged to identify certain seasonality and broader market trends at a given point in time, while simultaneously allowing for cross-deal comparison that normalizes for different issuance vintages. The table below demonstrates how Deal Age across different origination vintages is aligned under this approach for a hypothetical securitization with a cut-off date of January 15th, 2019 and a first collection period end date of February 28th, 2019.

Please note this view is only available for securitization datasets.

Loan Age Analysis

For Loan Age Analysis charts, data is normalized by loan age / months on book.  These views can be used to identify trends over the life of a given cohort, normalizing across origination vintages.  When using this view, it is important to be mindful of the fact that individual loans can have the same loan age at different points in time (i.e.; a different calendar month).  The table below illustrates this concept, showing how the numerator and denominator would be aggregated to calculate CPR at loan age 2 for a sample pool of loans originated in Q1’19.

dv01 calculates Loan Age using the expected first payment date at the point of origination. On the day where the borrower is expected to make his / her first payment, Loan Age will increment from 0 to 1. Where expected first payment date is not explicitly available, dv01 uses secondary logic to increment Loan Age one month from the origination date. In some cases, dv01 uses a Loan Age value explicitly reported in the raw data at the request of the originator / issuer.

Additional Information

Data Refresh Schedules

dv01 does not update data for any of the Analysis and Performance views without complete records being available. Below is a summary of how each type of view refreshes as new data is received:

• Historical Analysis: Updates on a monthly basis, only after a full calendar month of data has been received and processed.
• Deal Age Analysis: Updates on each deal’s distribution date, showing performance through the end of the previous collection period end date.
• Loan Age Analysis: Updates on a weekly basis, or when new data is received, only for loans that have completed a full loan age increment since the previous update.

Forced Vintage Grouping

dv01 enforces certain vintage grouping requirements on Historical and Loan Age charts to display consistent and logical results.  Non-static pools, such as portfolio and platform datasets, are constantly changing, as new loans are purchased, sold, or originated within any given month; therefore, any cumulative charts would not make sense without a forced vintage grouping, as the denominator would constantly be changing as loans enter and exit a given pool. dv01 defaults to a forced grouping of Vintage Quarter, but users can toggle this to use any other vintage grouping option available. All cumulative performance metrics leverage this forced grouping, including:

• Cumulative loss charts
• Cumulative prepayment charts
• Pool Factor
• ROI
• NAR

Please note, that non-cumulative performance metrics do not enforce this grouping.

For non-revolving securitizations, which are a fixed pool as of the transaction’s cut-off date, dv01 does not require forced vintage groupings for these calculations. Because these are fixed pools, the cumulative denominator does not change over time, and as a result, this requirement is not necessary.

Head Period Removal for Loan Age 

In addition to forcing vintage grouping requirements, dv01 also removes certain head period records for certain Loan Age views to eliminate incomplete records. 

Tail Period Removal for Loan Age 

In addition head period removal, dv01 also removes certain tail period records from the end of curves for certain Loan Age views to eliminate incomplete records.  For any performance metrics where a vintage grouping is used in the “Strat By” option, the following formula is used to determine the number of tail period records to be removed (max loan age to display):

n = the number of calendar months in a vintage grouping option
t = the max loan age available for a given cohort
max loan age displayed = t - (n - 1)

Similar to the rationale for forced vintage groupings, this logic is enforced to ensure that the denominator for any of the affected calculations remains static.

The table below demonstrates how tail-end loan age records would be removed for a hypothetical pool of loans originated in January, February, and March 2019 with performance data available through December 31st, 2019. Note how loan age records 10 and 11 are removed for January 2019 originations, while loan age record 10 is removed for February 2019 originations, to ensure that all data points displayed have a complete set of records associated with the corresponding Vintage Quarter grouping:

If the grouping was updated to represent Vintage Half-Year, the number of loan ages would be further reduced, only capturing records through loan age 6:

Please note, tail-end records are not removed for Historical and Deal Age Analysis views, as complete records exist for all data points.

The table below summarizes how the Forced Vintage Grouping and Tail Period Removal practices discussed above are applied to various dataset type / analysis type combinations for cumulative calculations:

Outlier Removal

There are often situations where performance can be erratic at the very tail end of a curve, when a given cohort dwindles down to a small percentage of its original size.  To address this, dv01 removes performance records where the count of loans associated with any given point across the x-axis on a chart is below 5% of the total count of loans in that same cohort, across time. This 5% threshold is applied after any grouping mechanisms are applied. For example, grouping performance of a securitization or portfolio by state, where many states will have far less 5% of the total loan count associated with the original pool, the trigger would only apply when a Loan Age or Historical period had less than 5% of the original loan count of any specific state.

There is a rare situation where Loan Age performance can be erratic even outside of tail periods, largely due to low loan counts. This is very rare, and it can occur in smaller securitizations or platforms composed of many different origination months. This behavior exists most often in securitizations, especially ones that are composed of seasoned loans, or are created by the collapse of multiple pre-existing transactions. When loans enter from multiple origination periods, some periods will have disparately few records, and, similar to the tail behavior discussed above, performance records in those periods may be misleading.