The dv01 cashflow engine projects future cashflows by simulating the performance of loans and waterfall structures based on user-specified assumptions.
The cashflow engine accepts static, step, and ramp vectors for Prepayment (CPR, SMM or ABS), Delinquency (% DQ per period), Default (CDR, MDR or ABS), and Loss Severity (percent). It also accepts a single, optional Severity Lag (in periods) and Forbearance Recovery. Curve assumptions can be anchored to the current period or anchored at origination by prefixing any curve notation with "AO:".
Order of Application
For each simulation period, the cashflow engine implements industry-standard application order:
- first apply default assumptions
- compute scheduled payments
- apply prepayment assumptions and severity amount
- servicing fees are computed on the balance that do not default within the period
Loan-level Operation and Recasting
The dv01 cashflow engine operates at loan-level with prepay and default assumptions applied as partial prepay and default to every loan in the cohort that the assumptions apply to.
Default assumptions specify immediate default: the proportion of loans that are affected by the period Default assumption immediately default and generate the specified amount of Severity. If an optional Severity Lag is specified, the Severity value cashflow is generated after the Lag number of periods. The proportion of loans that default do not generate servicing fees, including during the Lag periods.
Q: When applying a default curve (e.g. CDR) do the loans transition through the 30, 60, 90 delinquency states beforehand?
A: No, the default curves express a jump to default for two main reasons:
- in order to delay the recovery of these loans (which will be implemented via recovery lag)
- to delay the removal of the loan's principal from the pool's starting and ending balance, this would have implications for servicing)
Additionally, the dv01 cashflow engine operates at the loan-level with prepay and default assumptions applied as partial prepay and default to every loan in the cohort that the assumptions apply to.
Q: Do you support recovery lag?
Q: Off which balance is prepayment calculated?
A: starting balance - scheduled principal. In more detail, (DQ from previous month is paid back before this point, so starting balance would be reduced, but our DQ “repayment” is 0%.
Q: Off which balance is default calculated?
A: Starting balance. One defaults before they attempt to make their scheduled payment.
Q: Which balance is used to calculate servicing?
A: starting balance * gross interest / expected interest
If a loan payment wasn't made (it has defaulted), then the servicer does not receive payment.
Q: Which balance is used to calculate effective GWAC?
A: starting balance - defaulted balance – delinquent balance
Q: How is the monthly payment of a loan affected by partial prepayment and/or default?
A: In curtailment when a loan partially prepays (not in full/PIF), the same scheduled payment amount is expected the next month.
However, most cite CPR for the aggregate pool of loans rather than a partial prepayment of a single loan. Additionally, partial defaults don't exist and as a result, the monthly payment amount is reduced. This is standard practice with the partial prepayment/default approximation.
Q: How do I verify the expected interest/principal given a set of cashflows including GWAC?
A: You can't, the cashflows for a pool are the aggregate of many loans and the monthly payment is not equal among the many periods. To tie out the expected payments, you need to calculate it for each group of distinct remaining term or at loan level.
Using the current balance and GWAC metrics grouped by loan term remaining on the performance pages can expedite the process with the excel PMT function.
Q. Can step rates be modeled in future interest collections?
A. Yes, for data-sets that include step schedules, the loan coupon is incremented appropriately and interest is calculated off the updated coupon.
Q. How are forward curves modeled?
A. Forward rates are provided by ICE and updated daily ~4pm. Currently users cannot modify the forward curve (rate shocks) or upload their own curve assumptions.
Q. How are rates determined for ARMs?
A. The index type & forward curve (via ICE or Treasury) in conjunction with the specified lookback days/convention.
Q. What type of notations styles can I use for assumptions?
Q. How are cashflow assumptions applied? How does dv01 amortize loans?
A. Order of Application - For each simulation period, the cashflow engine implements industry-standard application order: it first applies any default assumptions, next any delinquent balances, then computes scheduled payments, and finally applies prepayment assumptions and any severity amount. Servicing fees are computed on the balance of the loans that do not default within the period.
Loan-level Operation and Recasting - The dv01 cashflow engine operates at loan-level, and thus prepay and default assumptions are applied as partial prepay and default to every loan in the cohort that the assumptions apply to.
Defaults - Default assumptions specify immediate default: the proportion of loans that are affected by the period Default assumption immediately default and generate the specified amount of Severity. If an optional Severity Lag is specified, the Severity value cashflow is generated after the Lag number of periods. The proportion of loans that default do not generate servicing fees, including during the Lag periods. See: Cashflow Calculations
Q. What are key differentiations from other external solutions (e.g.; Intex, Bloomberg)?
A. While the language results in similar functionality, the dv01 cashflow engine sources quality dv01 loan level data (not repline). Additionally, the dv01 cashflow engine supports benchmark and portfolio data-sets in addition to securitizations. Lastly, users can leverage an Anchor Orig notations style, AO:, to normalize assumptions by loan age.
Q. Does the cashflow engine support non-monthly pay loans, such as loans with 24 or 52 payments/year? How do they amortize?
A. dv01 supports sub monthly cashflows by using an approximation that there are an equal number of payments in each month. The payments for the month are calculated and the prepayment and default are applied to the aggregate monthly number (as CPR/CDR are annualized SMM and MDR - monthly numbers).
Q. Do you have a model?
A. dv01 features ADCo’s Loan Dynamics model for many mortgage datasets. Additionally, users can leverage Market Surveillance data-sets to create their own assumptions, outlined here: Run Cashflows Using CPR & CDR Curves from Loan Age Analysis