mortgagetape^™

Median lock rate + median LLPA (bps) per FICO × LTV cell over the trailing 14 days. Cells with fewer than 5 locks are suppressed. Conforming cells also carry an LLPA overlay-model band — a within-cell residual-risk rank from a HGB + isotonic model trained on 12M GSE acquisitions (2014-2020) with 2021-2022 held out (Phase 4 verdict NARROW). Bands: low · baseline · elevated · high. FHA/VA products don't show bands (model is GSE-trained). Computed with a national-representative loan profile anchored on TX; state-specific bands available via chat or the Beyond-the-grid article.

Composition

Trailing 7-day mix.

Geography

Conforming 30YR Fixed Purchase, trailing 14 days for OB; latest full year for GSE acquisitions (Fannie SFP + Freddie STACR, purchase only, deduped to one row per loan). OB − GSE positive = state is over-indexed in the current lock pipeline vs what closed agency last year; negative = under-indexed (share may be shifting to non-agency execution or the pipeline is rotating out). SFR rent YoY is the unweighted mean of Beekin single-family rent index YoY across the state's MSAs (latest month).

Rental signal

National single-family and multi-family rent index YoY, plus the metros with the strongest and weakest 12-month rent trajectory. Data from our Beekin rental income feed.

National = unweighted mean across all metros in the Beekin panel (August 2026 observation). Top/bottom-10 metros ranked by SFR rent index year-over-year change at the same observation. Beekin publishes the single-family and multi-family rent indices monthly at the MSA level for 887 metros back to 2015.

Pull-through & timing

Two different funnel views, presented side-by-side rather than differenced. OB post-lock funnel starts at lock request — median close→funded rate runs 85-92 % because most locks that reach a closing table actually fund. HMDA model App → origination (our pull-through classifier, AUC 0.93 on 2018-2023 HMDA) scores the whole application-to-origination journey, which is naturally 20-45 % — huge share of HMDA applications are withdrawn, denied, or never reach lock. The two aren't the same metric, so we don't compute a delta. What they do together: the OB column tells you today's close-out efficiency; the HMDA-model column anchors the industry-wide base rate for the same borrower profile.

LLPA reality check

Median observed LLPA (bps) minus the currently-effective Fannie base grid, per FICO × LTV cell. Positive = market above grid; negative = market below.

Conforming 30YR Fixed Purchase, trailing 14 days. Positive = market prices above the Fannie base grid; negative = market prices below. Deep dive: Beyond the grid.