There are three different ways to view model performance using the Dynamic Performance Report web site.  The first is by looking at performance across coupon buckets.  On this page, the user specifies the loan type (FNMA 30, GNMA 15, etc.), the origination year and quarter, the beginning and ending months of analysis, and any tuning values desired.  See figure 1 below.

figure 1

 

When not specified, the origination quarter and year fields default to “all,” which begins with data as far back as 1992.  When using the default value, the report uses weighted averages by remaining balance in the prepayment calculation.

 

For the months of analysis, the user can specify up to 10 years worth of monthly data to be used in the calculation.  When multiple months are used, the report again uses weighted averages by remaining balance.  The values given in the output grid (figure 2) for balance is the average monthly balance for that pool.  For example, when using an analysis period of Jan-03 to Apr-03, the balance reported for the 6.0 coupon bucket is what the average monthly balance was over that 4 month time period.

 

figure 2

 

 

The second view used to examine model performance is by viewing reports across origination years.  From this page, the user selects the loan type, the beginning and ending months of analysis, the coupon range desired, and any tuning values desired.

See figure 3 below.

figure 3

 

The first two inputs are the same as the previous one.  In this view, however, the user inputs the coupon or coupons to be used in the analysis.  When left blank, the report defaults to all coupons, which in the database is a range of 4.5 to 9.5, depending on the loan type chosen.  A user can specify any range of coupons.  To view 6.5’s, for example, the user would enter 6.5 in both inputs in the coupon range field.  If 6.5 is only entered in the first input, the report will use coupon buckets above and including 6.5.  Like the first report, the balances reported in the output table are the average monthly balances for that vintage across the analysis period specified.

 

The third view used on this site is by analysis month, which shows the model’s performance over a selected time period.  You can use this view to track a specific collateral from origination up to the current month.  The inputs here are loan type, origination quarter and year, the coupon or coupon range of interest, and the beginning and ending analysis months.  The rules for inputting the coupons are the same as for the performance by vintage view.  See figure 4 below.

figure 4

 

In this view, the balances listed in the output table are the actual balances in that month for the coupon or coupons selected from the input table.

 

TUNING

 

The tuning parameters available in these reports match those from the Prepayment Model in Excel and other vendor systems.  To view how each tuning parameter effects the model output, please refer to http://www.ad-co.com/support/newglossary.htm#tuning.  A user name and password is required.

 

In a previous issue of Andrew Davidson & Co.’s newsletter, The Pipeline, we discussed the recommendation of setting the burnout tuning parameter for the 30-yr models to 1.1.  This reduces the impact of burnout in the prepayment forecast, which results in faster speeds.  This recommendation is for FHLMC, FNMA, and GNMA 30-yr models.