Please ensure Javascript is enabled for purposes of website accessibility
top of page
  • Writer's pictureSharad Chaudhary

Mortgage Rate and Incentive Calculations in StoryBook

Conforming Mortgage Rates in StoryBook

The mortgage rates used in StoryBook are based on the most reliable and long-running survey of primary mortgage rates for conforming loans -- Freddie Mac's Primary Mortgage Market Survey (PMMS) (https://www.freddiemac.com/pmms/). Each week, from April 1971 until November 2022, PMMS was based on surveying lenders on the rates and discount/origination points for their most popular 30-year fixed-rate, 15-year fixed-rate, 5/1 Hybrid adjustable-rate, and 1-year adjustable-rate mortgage products. The survey was based on first-lien prime conventional (i.e., non-government) purchase mortgages with a conforming balance, an LTV of 80%, and excellent credit (FICO scores of 790 or above).


Legacy PMMS would report both the average rate and the average points each week. Since fees and points fluctuate over time, StoryBook normalizes rates to reflect a constant 1-point (discount + origination points) assumption. The normalization assumes that 1 point = 25bps for 30-year mortgages. For example, for the week of 10/27/22, the reported 30-year rate was 7.08%, with 0.80 points. The 1-point rate adjustment is then given by (0.80 – 1.00) * 25bps = -5bps, giving us 7.03% as the 1-point rate for that week.


Starting November 17th, 2022, Freddie Mac changed its process from surveying lenders to using the applications submitted by lenders to Freddie Mac's Loan Product Advisor (LPA) automated underwriting system when a borrower applies for a mortgage. (https://www.freddiemac.com/pmms/about-pmms). Only rates on 30-year and 15-year mortgages are published in the new PMMS and the fees/points associated with these quotes will no longer be published. The reference population of loans used to calculate averaged rates is very similar to the one used in the previous iteration of PMMS: conforming balance, single-family (only 1 dwelling unit), owner-occupied, 75%-80% OLTV, FICO scores above 740. The requirement that the loan be for a purchase mortgage has been dropped.


While Freddie Mac's research shows that the agreement between the new and legacy PMMS series is very close, the series are not identical and it is not possible to harmonize them given the lack of quoted points for the new series. However, in order to enforce some degree of consistency between the two series, StoryBook assumes that the quoted discount points/fees associated with the new series is always 0.8 points, which results in a 1-point 30-year rate that is consistently ~5bps lower than the published PMMS rate.


Closing Lags

The cycle time (or closing time) associated with a mortgage application is the time interval between initiating the application and funding the loan (assuming that the application is successful). The takeaway from this gap between mortgage application and closing is that incentive calculations need to be appropriately lagged in order to correctly align the prepayment rate observed in a particular month for a collection of loans with the incentive that these loans experienced (the borrower responds to the prevailing mortgage rate at the time of application).


StoryBook sources closing lags every month using data from ICE Mortgage Technology, which maintains one of the largest repoistories of loan-level residential mortgage data collected from its loan origination platform (https://www.icemortgagetechnology.com/) The same lag is assumed to hold across products and loan purposes for convenience since variations across these categories are typically small (less than 5 days).


Incentive Calculations

The incentive corresponding to any reporting month (say M) is calculated as the difference between a loan's note rate (or the weighted-average coupon (WAC), in the case of pools) and the lagged mortgage rate corresponding to M.

The recipe for calculating lagged mortgage rates is as follows. For simplicity's sake, we'll focus on the 30-year mortgage rate. The initial step is to convert the history of the weekly PMMS time series into a daily series by assuming that the 30-year mortgage rate is constant in the week covered by the PMMS. Denoting the lag associated to the reporting period M by L(M), we assume that the mortgage rate corresponding to each day in M is the daily mortgage rate from L(M) days ago. Then, the lagged mortgage rate for the entire reporting month is just the average of these daily rates, which are first normalized to 1-point rates before averaging, since the discount/origination points associated with these rates varies from week to week.


The incentive for each mortgage product is based on that product's mortgage rate. This of course is an approximation since borrowers do from time to time refinance across mortgage products: for example from a 30-year mortgage to a 5/1 Hybrid ARM or from a low LTV FHA mortgage to a conventional mortgage. The incentive calculation currently also does not take factors such as mortgage insurance premiums and loan-level pricing adjustments into account and thus will overstate the incentive for mortgages that have layers of credit risk present.


Mortgage Rates for Other Products

Unlike 30-year and 15-year conforming balance conventional loans, where the Freddie Mac PMMS rate is widely accepted as the benchmark mortgage rate series, there's no market consensus on the appropriate mortgage rate series to use for government loans (FHA/VA/USDA), conventional 30-year Jumbo loans etc. Consequently, we adopt the following approach to infer the historical mortgage rate series for such products.


The basic recipe is:

  1. For each month, use the loan-level MBS data sets released by the Agencies to construct reference populations for the 30-year conforming benchmark and "non-benchmarked" products that were originated in that month;

  2. Calculate the average note rate for each of these populations;

  3. Estimate the historical spread between the conforming benchmark and each of the non-benchmark populations for each month; and,

  4. Apply the relevant spread for each of the non-benchmark products to the current daily PMMS rate to estimate the prevailing daily mortgage rate for that product. Since loans are typically pooled with 1-2 months of seasoning, origination data lags current data by 1-2 months, so spreads for the most recent PMMS observations are generated by taking a moving average of historical spreads.


Why can't we directly aggregate the note rates on at-issue loan-level data to estimate non-benchmark rates? The problem, as noted in the last step of the recipe above, is that loans are typically placed in pools some months after they are originated. Thus, at-issue data collected in March will mostly reflect loans originated in January and February and thus only provide a snapshot of mortgage rates over that period, not as of the current month.


Reference populations are constructed by filtering the overall population of loans originated in a particular month (for a specific mortgage product) based on loan purpose, OLTV ranges, credit score ranges, and loan sizes. For example, for GSE 30-year conforming loans, we have the following assumptions:

  • Loan Purpose: Purchase

  • OLTV Range: 60-80%

  • FICO Range: 700-800

  • Loan Size: Conforming balance for the origination date


Similarly, for FHA 30-year loans:

  • Loan Purpose: Purchase

  • OLTV Range: 80-97%

  • FICO Range: 620-720

  • Loan Size: Satisfying FHA loan limit for the origination date


Customizing Calculations in StoryBook

As the above discussion demonstrates, calculating incentives and, by extension, S-Curves, involves several assumptions. StoryBook provides users with the flexibility to overlay their own assumptions:

  • An adjustment can be applied to the rate series and lags can be reset thus creating custom mortgage rates;

  • The impact of these changes can be visualized;

  • The changes can be propagated to all incentive calculations thus allowing users to regenerate S-Curves on the fly; and,

  • Users can also adjust the bucket-size used in calculating S-Curves.


We provide a visual guide to how these customizations can be achieved in StoryBook. Begin by clicking on the "Mtg Rates" button to get to the mortgage rate interface:



There are 11 different rate series for various products/terms, choose the one(s) of interest:




Enter your "Rate Adj/Lag/Inc.Bkt" assumption and click on "Explore" to see the difference between the default StoryBook series and your assumption. When you are satisfied with the series generated using your new assumptions, click on "Apply" -- incentives and S-Curves will now reflect these new assumptions.










62 views0 comments

Recent Posts

See All

Ginnie Mae Loan-level Data on BD4

Introduction In addition to providing pool-level A factors on GNMA MBS pools on the 4th business day (BD4) of each month, StoryBook also offers a loan-level data set that is derived from the loan liqu

Extended-Term (ET) Loans in GNMA Pools

Background An ET loan corresponds to a loan modification that lowers a borrower's monthly payments by restructuring an existing 30-year loan into a 40-year term (or some term between 30 and 40 years),

LLPA Changes and Mortgage Rates

LLPA Changes and Mortgage Rates On January 5th, 2022, the GSEs released a new round of loan-level pricing adjustments (LLPAs) targeted at conforming Jumbos and Second Homes. LLPA adjustments for jumbo

Comments


bottom of page