Every so often, current events deviate sharply from the past and the model needs to be adjusted to capture recent collateral performance more accurately. Adjustments can also be made to help the user assess the model’s sensitivity to various parameters or to reflect an opinion on the impact that various loan types and issuers will have on collateral performance.
As needed, AD&Co may issue tuning adjustment recommendations for a particular model and version. Tuning recommendations are issued by the respective model analyst/developer tracking the actual versus forecasted performance of the model on a monthly basis. Any new tuning recommendations will be announced in The Pipeline and will be available in the following table. While AD&Co advises that all tuning recommendations be employed, the user must decide if these tunings make sense for a given portfolio.
Vectored Tuning
Vectored tuning allows users to specify a time dimension to tuning parameters. Users can select the number of periods the tuning parameters are in effect. Please Note: If given a choice, when setting vectored tuning for any of the tuning parameters listed below, please select "Sensitivity" tuning as opposed to the "Model" tuning.
Tuning Definitions
Scale Using the scale tuning parameter simply adjust the model’s SMM vector by multiplying by the
value entered. For example, if the user wishes to speed up the model forecasts, they would use
a value over 1.0. This parameter does not adjust any of the underlying factors within the
model, it simply changes the final forecast value. Values for scale can be any number greater
than 0, and the default value is 1.0.
Refi
The refi tuning parameter also works as a multiplier. It affects the Refi_SMM component of
the overall model SMM. By adjusting this parameter, the user can increase the amount of
SMM forecast by the model at a given incentive level. Internally, the model calculates a
refinance incentive for a loan/pool based on a number of characteristics, and there is an
estimated SMM level at each level of incentive based off the S-curve within the model.
Adjusting the refi tuning parameter above 1.0 will increase SMM forecasts for those
loans/pools with any level of refinance incentive, and setting to below 1.0 will slow down
forecasts for those same loans/pools. This is a useful parameter to use if the model is
underpredicting or overpredicting prepayments for premiums, as it can affect those loan/pools
without changing the values for discounts. Values for refi can be any number greater than 0,
and the default value is 1.0.
Turnover
The turnover tuning parameter is a multiplier on the Turnover_SMM component of the
model. Changing this value will raise or lower the level of basic housing turnover forecasted
by the model. This will affect all loans/pools, regardless of refinance incentive. Using the
turnover tuning parameter can be helpful whenever there are temporary housing spikes or
slowdowns, allowing the model to reflect short-term changes. Values for turnover can be any
number greater than 0, and the default value is 1.0.
Cashout
Tuning the cashout tuning parameter allows the user to strengthen or weaken the amount of
Cashout_SMM in the model forecast. This piece of the model is designed to capture the
additional amount of prepayments seen from home price appreciation, where borrowers cash
out the equity in their homes by refinancing into a new loan. The model will forecast a given
amount of cashout_SMM based on the two-year change of the national house price index. For
example, the model might forecast an additional 3 CPR for a two-year home price change of
12%, but tuning the cashout parameter will allow the user to increase or decrease that CPR for
the given level of 12% appreciation. This parameter also works as a multiplier, and the values
can be any number greater than 0, and the default value is 1.0.
Credit Cure
The credit cure tuning parameter functions in the same way that cashout does. It is a
multiplier directly on the Credit_Cure_SMM forecasted by the model. The credit cure
function attempts to forecast prepayments by a borrower who has seen a rise in their creditworthiness,
and can refinance a loan to a better interest rate. This function is heavily
dependent on the age of the pool/loan, so the credit cure tuning parameter allows the user to
increase or decrease the amount of SMM the model forecasts at any given age of the
pool/loan. Values for credit cure can be any number greater than 0, and the default value is
1.0.
Slide
The slide tuning parameter is different than all the other tuning parameters. It is designed to
change the refinancing S-curve to make the model more or less sensitive to prepayments at
different incentives. By adjusting the slide parameter, the user can change the level of refi
incentive where prepayments begin to increase. The units of slide are in basis points, and any
values entered modify the underlying mortgage coupon for the model being used. For
example, if a user is using the FNMA 30YR model, and tunes slide to 25, the user is
essentially adding 25 basis points to the FNCL, the underlying mortgage rate the FNMA
30YR model uses to calculate refinance incentive. Adding basis points to the coupon “slides”
the S-curve to the right, meaning prepayments will start to increase at a higher coupon level,
and subtracting basis points from the coupon “slides” the S-curve to the left, meaning
prepayments will start to increase and a lower coupon level. This is most useful for applying a
model to loans/pools it was not estimated on, for example like agency quality loans that little
or no documentation. Increasing slide for these loans will help reflect the different levels of
interest rates available for these borrowers. Values of slide should be in the range of -500 to
500, and the default value is 0.
SATO
The SATO tuning parameter affects the spread at origination effect that is built into the
model. This effect dampens prepayments for loans/pools that have a high coupon relative to
interest rates at the time the loan/pool was originated. This effect is in place for a finite
amount of time from the origination of the pool/loan, and the tuning parameter weakens or
strengthens the dampening that occurs during that time. For example, if the user tunes SATO
to 0, this effectively turns off the SATO effect, meaning that newly originated premiums will
have the full amount of SMM projections from age 1 going forward. Values for SATO can be
any number greater than 0, and the default value is 1.0.
Curve Spread
Curve spread tuning adjusts the model factor that controls for increased or decreased refi
incentive based on the spread of the yield curve. If the spread is wide, there is added incentive
to refinance from fixed to ARM, and less incentive to refi from ARM to fixed. If the curve is
flat, there is less incentive to go from fixed to ARM. The curve spread tuning factor is a
multiplier directly on this effect; it increases or decrease the prepayment speed based on the
type of pool/loan being analyzed. Values for curve spread can be any number greater than 0,
and the default value is 1.0.
Aging
The aging tuning parameter is designed to speed up or slow down the aging effect in the
model. The aging effect controls when the model forecasts loan/pools to reach their full
amount of turnover and refi SMM’s. Increasing the aging parameter will make the model
reach the full amount of turnover and refi SMM sooner, and decreasing it towards 0 will push
the aging peak out further. Values for aging can be any number greater than 0, and the default
value is 1.0.
Lag
The lag tuning parameter allows the user to adjust the lag weighting on the mortgage coupon
used when determining refinance incentive. The model uses a blended mortgage interest rate
from the previous 3 months to determine the relative refinance rate for a given loan/pool.
Adjusting the lag parameter above 1.0 puts more weight on the interest rate from 3 months
ago, and decreasing towards 0 places more weight on the rate from last month. Values for lag
can be any number greater than 0, and the default value is 1.0.
Burnout
Burnout tuning allows the user to adjust the rate at which the model’s burnout occurs. There
are two types of burnout tuning, “model” and “sensitivity”. Choosing “model” will change the
historical burnout rate (if you are running an analysis where the loan/pool has already aged)
from pool/loan origination, and going forward into forecast. Choosing “sensitivity” will only
change the burnout rate going forward from today, i.e. in the forecast only. This tuning
parameter is useful for situations where the user feels that the model does not slow down
enough when the loan/pool becomes seasoned, or where too much prepayment occurs too soon
and the speeds being forecast for the older ages are too low. Increasing this parameter above
1.0 will make more prepayments occur sooner, and decreasing towards 0 will make seasoned
collateral prepay at a higher rate. Values for burnout can be any number greater than 0, and
the default value is 1.0.
Tuning Support
Prepayment Tuning Recommendations
Tuning Release Date
Model Version Affected
Tuning Recommendation
Every so often, current events deviate sharply from the past and the model needs to be adjusted to capture recent collateral performance more accurately. Adjustments can also be made to help the user assess the model’s sensitivity to various parameters or to reflect an opinion on the impact that various loan types and issuers will have on collateral performance.Why Tune the Model?
As needed, AD&Co may issue tuning adjustment recommendations for a particular model and version. Tuning recommendations are issued by the respective model analyst/developer tracking the actual versus forecasted performance of the model on a monthly basis. Any new tuning recommendations will be announced in The Pipeline and will be available in the following table. While AD&Co advises that all tuning recommendations be employed, the user must decide if these tunings make sense for a given portfolio.
Vectored Tuning
Vectored tuning allows users to specify a time dimension to tuning parameters. Users can select the number of periods the tuning parameters are in effect. Please Note: If given a choice, when setting vectored tuning for any of the tuning parameters listed below, please select "Sensitivity" tuning as opposed to the "Model" tuning.
Tuning Definitions
Scale
Using the scale tuning parameter simply adjust the model’s SMM vector by multiplying by the
value entered. For example, if the user wishes to speed up the model forecasts, they would use
a value over 1.0. This parameter does not adjust any of the underlying factors within the
model, it simply changes the final forecast value. Values for scale can be any number greater
than 0, and the default value is 1.0.
Refi
The refi tuning parameter also works as a multiplier. It affects the Refi_SMM component of
the overall model SMM. By adjusting this parameter, the user can increase the amount of
SMM forecast by the model at a given incentive level. Internally, the model calculates a
refinance incentive for a loan/pool based on a number of characteristics, and there is an
estimated SMM level at each level of incentive based off the S-curve within the model.
Adjusting the refi tuning parameter above 1.0 will increase SMM forecasts for those
loans/pools with any level of refinance incentive, and setting to below 1.0 will slow down
forecasts for those same loans/pools. This is a useful parameter to use if the model is
underpredicting or overpredicting prepayments for premiums, as it can affect those loan/pools
without changing the values for discounts. Values for refi can be any number greater than 0,
and the default value is 1.0.
Turnover
The turnover tuning parameter is a multiplier on the Turnover_SMM component of the
model. Changing this value will raise or lower the level of basic housing turnover forecasted
by the model. This will affect all loans/pools, regardless of refinance incentive. Using the
turnover tuning parameter can be helpful whenever there are temporary housing spikes or
slowdowns, allowing the model to reflect short-term changes. Values for turnover can be any
number greater than 0, and the default value is 1.0.
Cashout
Tuning the cashout tuning parameter allows the user to strengthen or weaken the amount of
Cashout_SMM in the model forecast. This piece of the model is designed to capture the
additional amount of prepayments seen from home price appreciation, where borrowers cash
out the equity in their homes by refinancing into a new loan. The model will forecast a given
amount of cashout_SMM based on the two-year change of the national house price index. For
example, the model might forecast an additional 3 CPR for a two-year home price change of
12%, but tuning the cashout parameter will allow the user to increase or decrease that CPR for
the given level of 12% appreciation. This parameter also works as a multiplier, and the values
can be any number greater than 0, and the default value is 1.0.
Credit Cure
The credit cure tuning parameter functions in the same way that cashout does. It is a
multiplier directly on the Credit_Cure_SMM forecasted by the model. The credit cure
function attempts to forecast prepayments by a borrower who has seen a rise in their creditworthiness,
and can refinance a loan to a better interest rate. This function is heavily
dependent on the age of the pool/loan, so the credit cure tuning parameter allows the user to
increase or decrease the amount of SMM the model forecasts at any given age of the
pool/loan. Values for credit cure can be any number greater than 0, and the default value is
1.0.
Slide
The slide tuning parameter is different than all the other tuning parameters. It is designed to
change the refinancing S-curve to make the model more or less sensitive to prepayments at
different incentives. By adjusting the slide parameter, the user can change the level of refi
incentive where prepayments begin to increase. The units of slide are in basis points, and any
values entered modify the underlying mortgage coupon for the model being used. For
example, if a user is using the FNMA 30YR model, and tunes slide to 25, the user is
essentially adding 25 basis points to the FNCL, the underlying mortgage rate the FNMA
30YR model uses to calculate refinance incentive. Adding basis points to the coupon “slides”
the S-curve to the right, meaning prepayments will start to increase at a higher coupon level,
and subtracting basis points from the coupon “slides” the S-curve to the left, meaning
prepayments will start to increase and a lower coupon level. This is most useful for applying a
model to loans/pools it was not estimated on, for example like agency quality loans that little
or no documentation. Increasing slide for these loans will help reflect the different levels of
interest rates available for these borrowers. Values of slide should be in the range of -500 to
500, and the default value is 0.
SATO
The SATO tuning parameter affects the spread at origination effect that is built into the
model. This effect dampens prepayments for loans/pools that have a high coupon relative to
interest rates at the time the loan/pool was originated. This effect is in place for a finite
amount of time from the origination of the pool/loan, and the tuning parameter weakens or
strengthens the dampening that occurs during that time. For example, if the user tunes SATO
to 0, this effectively turns off the SATO effect, meaning that newly originated premiums will
have the full amount of SMM projections from age 1 going forward. Values for SATO can be
any number greater than 0, and the default value is 1.0.
Curve Spread
Curve spread tuning adjusts the model factor that controls for increased or decreased refi
incentive based on the spread of the yield curve. If the spread is wide, there is added incentive
to refinance from fixed to ARM, and less incentive to refi from ARM to fixed. If the curve is
flat, there is less incentive to go from fixed to ARM. The curve spread tuning factor is a
multiplier directly on this effect; it increases or decrease the prepayment speed based on the
type of pool/loan being analyzed. Values for curve spread can be any number greater than 0,
and the default value is 1.0.
Aging
The aging tuning parameter is designed to speed up or slow down the aging effect in the
model. The aging effect controls when the model forecasts loan/pools to reach their full
amount of turnover and refi SMM’s. Increasing the aging parameter will make the model
reach the full amount of turnover and refi SMM sooner, and decreasing it towards 0 will push
the aging peak out further. Values for aging can be any number greater than 0, and the default
value is 1.0.
Lag
The lag tuning parameter allows the user to adjust the lag weighting on the mortgage coupon
used when determining refinance incentive. The model uses a blended mortgage interest rate
from the previous 3 months to determine the relative refinance rate for a given loan/pool.
Adjusting the lag parameter above 1.0 puts more weight on the interest rate from 3 months
ago, and decreasing towards 0 places more weight on the rate from last month. Values for lag
can be any number greater than 0, and the default value is 1.0.
Burnout
Burnout tuning allows the user to adjust the rate at which the model’s burnout occurs. There
are two types of burnout tuning, “model” and “sensitivity”. Choosing “model” will change the
historical burnout rate (if you are running an analysis where the loan/pool has already aged)
from pool/loan origination, and going forward into forecast. Choosing “sensitivity” will only
change the burnout rate going forward from today, i.e. in the forecast only. This tuning
parameter is useful for situations where the user feels that the model does not slow down
enough when the loan/pool becomes seasoned, or where too much prepayment occurs too soon
and the speeds being forecast for the older ages are too low. Increasing this parameter above
1.0 will make more prepayments occur sooner, and decreasing towards 0 will make seasoned
collateral prepay at a higher rate. Values for burnout can be any number greater than 0, and
the default value is 1.0.