Rates vs. Prices: Which Variable Has Driven US Housing Affordability More Since 1971? A 54-Year Dataset
- Since 1971, a 100 basis-point rise in the 30-year mortgage rate has the same monthly-payment impact as roughly a 10% increase in the home price. The two levers are not equally weighted: at the median, the rate channel explains 74% of the variation in monthly payments; prices account for the remaining 26%.
- This rate dominance is not a statistical artifact of the Volcker episode. Removing October 1979 through December 1982 from the sample still yields a 70% / 30% rate-vs-price split. In 5 of the 6 decades covered, rates explain more than two-thirds of payment variation — the 2000s are the single exception (54% / 46%), the decade of the housing bubble.
- As of April 2026, the 30-year mortgage rate stands at 6.38%, with the median new-home sales price at $410,800 and median family income at $105,800. The implied monthly principal-and-interest payment on an 80% LTV loan is $2,051, or a 23.3% payment-to-income ratio (PTI) — the 59th percentile of all monthly observations since 1971.
- The historical PTI extremes are 45.0% in October 1981 (mortgage rate 18.45%, Volcker peak) and 14.9% in December 2020 (mortgage rate 2.68%, pandemic low). The recent cyclical peak — October 2023 at 27.5% PTI — is well below the 1981 benchmark despite nominal home prices being roughly 6× higher.
Which variable has driven US housing affordability more since 1971? A variance decomposition
US housing affordability is the product of three moving parts — the mortgage rate, the home price, and household income — combined through a 30-year amortization formula. When commentators argue that “houses are unaffordable,” they implicitly take a position on which variable matters most. This study runs that decomposition explicitly, across 661 monthly observations from April 1971 through April 2026.
The result is asymmetric. Although both the mortgage rate and the home price enter the monthly-payment calculation directly, the rate has swung across a ~7× range (from 2.68% to 18.45%) while the home price has moved by about 16× in nominal terms. The arithmetic asymmetry matters less than one might expect: rates operate on the payment through the annuity factor, which is highly non-linear in the rate level. A single 100bp move at 4% has a larger proportional payment effect than the same 100bp at 12%, but across the full historical distribution the rate channel still explains about three-quarters of year-over-year payment variation.
The exception is the 2000s housing bubble, when prices rose fast enough to temporarily match the pass-through of the Fed’s rate cycle. That episode, however, remains the single decade in 54 years in which the price channel approached parity with the rate channel. In every other decade covered — including the 2020s — rates have dominated.
How the monthly payment is calculated
- Mortgage rate: Freddie Mac Primary Mortgage Market Survey (weekly), 30-year conventional fixed, retrieved from FRED (series MORTGAGE30US). Weekly observations are aggregated to a monthly mean.
- Home price: US Census Bureau and US Department of Housing and Urban Development, Median Sales Price of Houses Sold for the United States (MSPUS, quarterly), forward-filled to monthly frequency. This is the new-home median; existing-home medians are typically lower but track directionally.
- Income: US Census Bureau, Median Family Income in the United States (MEFAINUSA646N, annual), linearly interpolated to monthly frequency. This is the family, not the household, median — family income skews higher because it excludes single-person households.
- Loan assumption: 80% loan-to-value on the median home price, 30-year fixed amortization.
- Monthly payment: standard amortization formula —
M = P · r · (1+r)^n / ((1+r)^n − 1)— where P is the loan principal, r is the monthly rate, and n = 360. This is principal and interest only. Property tax, homeowner insurance, HOA dues, and private mortgage insurance are excluded — these vary widely by geography and loan structure, and adding them would obscure the rate-vs-price decomposition that is the focus of this study. - Payment-to-income ratio (PTI): annual payment / annual median family income, expressed in percent.
Two simplifications are worth flagging explicitly. First, the use of median family income rather than median household income — the latter is only available from 1984, and switching series would sacrifice more than a decade of coverage. Family income is systematically higher than household income (median family income in 2024 was $105,800 vs. $83,730 for households), so the PTI values reported here are roughly 20–25% lower than they would be under a household-income denominator. The rank ordering across time is preserved; the absolute levels are not directly comparable to a PTI computed on household income.
Second, MSPUS measures the price of new single-family homes sold during the quarter, not the price of existing homes. Existing-home medians (as tracked by the National Association of Realtors) are generally lower — closer to $400,000 in 2024 against an MSPUS reading above $415,000 — but both series move together cyclically. The rate-vs-price variance decomposition shown below is insensitive to which of the two price series is used; the directional conclusion is the same.
The rate channel explains 74% of payment variation
The monthly payment M can be decomposed multiplicatively into two factors: the annuity factor A(r) = r(1+r)^n / ((1+r)^n − 1), which depends only on the mortgage rate, and the loan principal P, which moves one-for-one with the home price under a constant LTV assumption. Taking natural logs:
This identity holds exactly. The variance of the log change in the monthly payment can then be written as the sum of the variances of the two components plus twice their covariance. Attributing the explained variance is done via a covariance-based decomposition — Cov(Δ log M, Δ log A) divided by Var(Δ log M) — which sums to 100% by construction and avoids the ambiguity of the raw-variance approach when the two channels co-move.
(1972–2026, n=649)
(excl. Oct 1979 – Dec 1982, n=610)
The standard quant objection to this finding is that the 1979–1982 Volcker disinflation dominates the rate-channel variance, inflating the result. The ex-Volcker panel addresses that directly: with the Volcker window excluded, the rate channel still accounts for roughly 70% of payment variation. The finding is robust.
The by-decade decomposition (Chart 3 below) extends this robustness check. The rate share is 73% in the 1970s, 98% in the 1980s (the Volcker-and-aftermath decade), 87% in the 1990s, 54% in the 2000s, 73% in the 2010s, and 73% in the 2020s-to-date. The 2000s stand out: this is the single decade in which price-channel variance approached parity with the rate channel, driven by the 2002–2006 price run-up and the 2007–2009 reversal.
Reading the matrix — the isoquants are the mechanism
The white contour lines are the visual expression of the amortization formula. Each line traces all (rate, price) combinations that produce the same monthly payment. The lines curve upward and to the left, because at low rates the annuity factor is small and the payment is dominated by the principal; at high rates the annuity factor inflates and the payment is dominated by the rate.
The 1981 Volcker peak and the 2020 pandemic low are not on the same isoquant: October 1981 sits on the $870/month line (at a $70,400 home price, 18.45% rate), while December 2020 sits on the $1,096/month line (at $338,600, 2.68%). The 2020 payment in nominal dollars was higher than 1981, despite rates being about one-seventh as high — because the home price was almost five times larger. The PTI comparison is the reverse (14.9% in 2020 versus 45.0% in 1981), because median family income had risen from $23,170 in 1981 to $88,237 in 2020.
The April 2026 point — the red star — sits on the $2,051/month isoquant: a $410,800 median new-home price at a 6.38% rate, producing a 23.3% PTI against median family income of $105,800. This is the 59th percentile of all 661 monthly observations — above the long-run median of 22.5% but well below the 75th percentile threshold (25.4%) that defines the “stressed” regime in the quartile-based classification used in Chart 1.
Rate-equivalent price impact — the 100bp-to-10%-price conversion
A convenient way to summarize the rate/price sensitivity is to ask: by how much would the home price have to rise to produce the same monthly-payment increase as a 100-basis-point rise in the rate? The answer depends on the starting rate, because the annuity factor is non-linear. Across all 661 monthly observations, the mean rate-equivalent price impact is 10.1%, with a median of 10.2%. The range is 5.3% (when rates are at their historical peak) to 13.5% (when rates are at their trough).
| Starting rate | Payment per $100K principal | +100bp impact on payment | Equivalent price increase (%) |
|---|---|---|---|
| 3.0% | $422 | +$56 | +13.2% |
| 4.0% | $477 | +$59 | +12.4% |
| 5.0% | $537 | +$63 | +11.7% |
| 6.0% (current) | $600 | +$66 | +11.0% |
| 7.0% | $665 | +$68 | +10.3% |
| 10.0% | $878 | +$75 | +8.5% |
| 15.0% | $1,264 | +$80 | +6.4% |
| 18.0% | $1,507 | +$82 | +5.4% |
At the current 6.38% rate, a hypothetical return to the October 2023 cyclical peak (7.62%, a +124bp move from today’s rate) would — holding prices constant — raise the monthly payment by about 13%, equivalent to a $55,000 increase in the median home price at today’s rate. Conversely, a 200bp cut would lower the monthly payment by about 20%, equivalent to roughly an $82,000 price reduction. The rate lever is simply much larger than any plausible short-run move in the home price.
Four reference episodes
The historical extremes can be compared in a single table. Each row is a monthly observation taken at a recognizable turning point. The PTI is computed consistently using the same methodology across all rows.
| Month | Rate | Median home price | Median family income | Monthly P&I | PTI |
|---|---|---|---|---|---|
| Oct 1981 — Volcker peak | 18.45% | $70,400 | $23,170 | $870 | 45.0% |
| Jun 2006 — housing bubble peak PTI | 6.68% | $246,300 | $59,640 | $1,269 | 25.5% |
| Dec 2020 — pandemic rate low | 2.68% | $338,600 | $88,240 | $1,096 | 14.9% |
| Oct 2023 — recent affordability peak | 7.62% | $423,200 | $104,550 | $2,395 | 27.5% |
| Apr 2026 — current | 6.38% | $410,800 | $105,800 | $2,051 | 23.3% |
rates-vs-prices-affordability.csv.Two features of the table are worth highlighting. First, the nominal monthly payment in October 2023 ($2,395) was larger than in October 1981 ($870), despite the 1981 rate being nearly three times higher. The compounding of 54 years of price inflation does the work. Second, the PTI in 2023 (27.5%) was far below the 1981 PTI (45%), because income had grown even faster than home prices over the same span. “Affordability” — as measured by PTI — and “payment in dollars” tell different stories.
What this decomposition does not measure
1. Property tax, insurance, and PMI are excluded. A realistic all-in monthly housing cost for a median US home today is closer to $2,500–$2,700 once property tax (nationally averaging around 1.1% of home value annually) and homeowner insurance (typically $1,500–$2,500/year) are added, with private mortgage insurance adding further to loans above 80% LTV. For a median family income of $105,800 the all-in PTI is closer to 29–31% — at or above the 28% guideline that most lenders and the Consumer Financial Protection Bureau flag as a stress threshold. The decomposition presented here is about the mechanical contribution of rates vs. prices to the P&I component; it is not a full affordability measure.
2. Regional variation dwarfs the national median. Median new-home prices in the San Francisco Bay Area, Los Angeles, and Seattle metros are roughly 2× the national MSPUS; median prices in several Midwest and Southern metros are 0.6–0.8× national. The same 6.38% mortgage rate produces a very different affordability profile in Detroit versus San Jose. The national decomposition should not be read as representative of any particular local market.
3. The 30-year fixed mortgage is an American institutional artifact. In most other developed markets, residential mortgages reset more frequently (e.g., UK 2–5-year fixed, France variable, Australia variable-dominant). In those systems the rate channel passes through to household budgets within months rather than decades. In the US, only new borrowers and refinancers feel a rate change — existing fixed-rate mortgagees are locked in. This “lock-in effect” has been widely documented in 2023–2024 housing-market commentary and is an important mechanical reason why the US housing market is not a simple real-time function of the Fed Funds rate.
The strongest case against this framing
The decomposition presented here can be challenged on two grounds that are worth stating in their strongest form.
First, the variance decomposition treats rates and prices as independent inputs. They are not. Mortgage rates are themselves an input into the home-price formation process: low rates pull forward housing demand and contribute to the price increases observed in the 2003–2006 and 2020–2022 episodes. If the “true” causal structure is that rates drive prices through a demand channel, then the 54% / 46% split observed in the 2000s understates the rate channel — a portion of what appears in the price column is also a rate effect working through the demand function. Under this interpretation, the reported 74% full-period rate share is a lower bound on the total rate influence.
Second, the exclusion of property tax, insurance, and PMI is analytically clean but practically misleading. These items have risen faster than general inflation in many US metros over the 2018–2025 period — Florida insurance premiums, for example, have more than doubled in several counties — and they are increasingly the binding affordability constraint for median buyers. A study that stops at P&I may understate the severity of the current affordability episode relative to the historical record, because the “non-mortgage” housing cost share has grown. The appropriate response is to publish the P&I decomposition (as this study does) alongside clear acknowledgment that it is a partial picture — which is the purpose of the caveats section above.
Frequently asked questions
Why use median family income rather than median household income?
The FRED series MEFAINUSA646N (family income) goes back to 1953; MEHOINUSA646N (household income) only goes back to 1984. Using family income preserves 13 additional years of history — including the Volcker period, which is analytically central. The downside is that the PTI levels reported are about 20–25% lower than they would be if household income were used (median family income in 2024 was $105,800 vs. $83,730 for households). The rank ordering of PTI across time is essentially unchanged; only the absolute levels shift.
Does the 2000s result imply that home prices are “just as important as rates”?
Only for that specific decade. The 2000s are the single decade in 54 years in which the price channel approached parity with the rate channel. In every other decade — 1970s, 1980s, 1990s, 2010s, 2020s — the rate channel has dominated, with shares of at least 73%. The 2000s result reflects the exceptional scale of the 2002–2006 price run-up followed by the 2007–2009 reversal, not a steady-state feature of the US housing market.
Why exclude property tax, insurance, and PMI from the payment?
For analytical clarity. This study is about the mechanical relationship between mortgage rates, home prices, and the principal-and-interest portion of the monthly payment. Adding property tax, insurance, and PMI would blur the rate-vs-price decomposition — those items depend on geography, loan structure, and insurance market conditions that are independent of the rate/price channels under study. The all-in monthly housing cost for a median US home is closer to $2,500–$2,700 than to the $2,051 reported here; that gap is the non-P&I component. The decomposition presented here is internally consistent but partial, and the caveats section is explicit about this.
Why is the 30-year fixed mortgage specifically relevant to this decomposition?
The 30-year fixed rate is an American institutional structure that does not exist in most other developed markets. Under this contract, only new borrowers and refinancers are exposed to the current market rate; existing mortgagees are locked into the rate they originated at. As of 2024, roughly 60% of outstanding US mortgage balances were at rates below 4% — reflecting originations from the 2020–2021 refinancing wave. This lock-in means that the market-clearing rate affects housing-market transaction volumes (depressed since 2022) more than it affects the aggregate household-sector mortgage debt service ratio. The decomposition in this study speaks to the marginal buyer, not the average US homeowner.
Can this methodology be applied to existing-home prices rather than new-home prices?
Yes, with caveats. The National Association of Realtors publishes median existing-home sales prices monthly from 1989. Using that series in place of MSPUS changes the absolute price level (existing-home medians are typically 3–5% lower than new-home medians) but leaves the rate-vs-price variance decomposition essentially unchanged — both series move together cyclically and the common component of their variation is what the decomposition captures. The pre-1989 portion of the study would have to be truncated, losing the Volcker period, which is the most informative single episode for identifying the rate channel’s power.
What would the decomposition look like if rates drove prices (reverse causation)?
Under a demand-side causal structure in which low rates pull forward housing demand and thereby raise prices, a portion of the measured price-channel variance would “really” be a lagged rate effect. The reported 74% rate-channel share would then be a lower bound on the total rate influence. The empirical identification of that demand-channel effect is a separate exercise — typically addressed in the academic literature via vector autoregression or instrumental-variables identification — and is outside the scope of this descriptive decomposition. The result reported here should be read as: “of the year-over-year variation in the median US monthly mortgage payment, roughly three-quarters is directly attributable to rate movement, holding price constant, and roughly one-quarter is directly attributable to price movement, holding rate constant.” Any indirect rate-through-price channel would be additional.
Dataset — reproducibility
The full monthly dataset underlying this study is published in open CSV format. All statistics in this page are reproducible directly from the CSV using standard spreadsheet software or any data-analysis tool.
date, mortgage_rate, median_home_price, median_income, loan_principal, monthly_payment, annual_payment, payment_to_income_pct, affordability_regime, rate_equiv_price_pct_per_100bp, volcker_flag, rate_yoy_delta, price_yoy_pct, payment_yoy_pct, dlog_annuity_yoy, dlog_principal_yoy, dlog_payment_yoy.https://eco3min.fr/en/rates-vs-prices-us-housing-affordability/. Underlying sources: Freddie Mac Primary Mortgage Market Survey via FRED (MORTGAGE30US); US Census Bureau and US Department of Housing and Urban Development, MSPUS; US Census Bureau, MEFAINUSA646N.
Related Eco3min studies and datasets
Last updated — 19 April 2026
Disclaimer – Financial Information: The analyses, commentary, and content published on eco3min.fr are provided for informational and educational purposes only. They do not constitute investment advice or a solicitation to buy or sell financial instruments. Past performance is not indicative of future results. All investment decisions involve risk and are the sole responsibility of the reader.