We take the upper bound of the 5-year mortgage rate forecast (the p90 band from the ensemble) and re-run DSCR on every row in every surface — holding rent and home value constant. The question: how many properties that pencil today stop penciling at the upper rate? This is a worst-case static stress; see the methodology block for why reality may be milder.
| Surface | Total rows | With rent + hv | Above 1.0 now | Above 1.0 stressed | Dropouts | Dropout % | Median DSCR now | Median DSCR stressed |
|---|
For every row in every surface that has both rent and home value, we
recompute the methodology DSCR with the rate replaced by the 5-year
upper-band rate — the p90 of the 4-model ensemble forecast at
year+5. Rent, home value, vacancy, taxes,
management, insurance, LTV, amortization — all held constant. Only
the rate moves.
A dropout is a row that has DSCR ≥ 1.0 at today's rate and DSCR < 1.0 at the upper-band rate. Rows that were already underwater don't count as new dropouts. Rows with null rent or hv stay null in both columns (no extrapolation).
Bond and housing markets often move together in real cycles. When mortgage rates spike, affordability collapses and home prices typically soften too — sometimes a lot. If hv falls along with rates rising, debt service drops, and the dropout count would be lower than what we show here.
This page deliberately freezes hv to isolate the rate channel. Read the numbers as "if prices held perfectly steady at today's level while rates drifted to the p90 of our 5-year forecast, this is what would happen." Reality almost never delivers that exact scenario. Use these numbers as an upper bound on rate-driven DSCR damage, not a forecast.
One rate number is applied identically across all four surfaces. No
per-surface rate adjustments, no per-state risk premiums, no LTV
stretching. The same RATE_NOW and the same
RATE_UPPER_5Y go through every row. This is the only way
cross-surface comparisons are honest.