Economics textbooks have a tidy answer for why interest rates matter. What counts, students learn, is the real interest rate, meaning the nominal rate minus expected inflation. If inflation runs at 5% and the mortgage rate is 8%, the “true” cost of borrowing is 3%. The nominal number on the loan document is, in theory, a veil.
Homebuyers, car shoppers, and anyone who has recently applied for a mortgage might raise an eyebrow at that. A new NBER working paper argues that their instinct is right and the textbooks are incomplete. The authors show that nominal rates have real effects on what households can afford, independent of what real rates are doing, and they build a model to trace out the consequences for the broader economy.
A question macroeconomics tends to skip
The paper is the work of Joshua K. Hausman of the University of Michigan, the late John V. Leahy, also of Michigan, John Mondragon of the Federal Reserve Bank of San Francisco, and Johannes Wieland of the University of California, San Diego. Their starting point is a mismatch between standard macroeconomic models and how debt actually works in the real world.
In most models, debt is a one-period contract settled in real terms, so only the real rate matters. But residential mortgages, car loans, and corporate bonds typically require a sequence of fixed nominal payments stretching out over years or decades. When nominal rates rise alongside inflation, those fixed payments get front-loaded in real terms: the early years feel heavier, the later years lighter.
That tilt collides with a second feature of real-world lending. Banks, whether because of regulation or fear of default, won’t approve loans where the initial monthly payment eats too large a share of a borrower’s current income. Borrowers themselves often refuse to stretch that far. The authors call this the payment-to-income constraint, and they argue it is the mechanism through which nominal rates sneak back into the real economy.
The arithmetic of a tilted payment
To see the idea, consider a household earning $100,000 a year that wants to borrow as much as a lender will allow, with payments capped at 50% of income. If expected inflation is zero and the real rate is zero, the nominal rate is also zero. A 30-year mortgage at 0% lets the household borrow $1.5 million, with payments of $50,000 a year that stay flat in real terms.
Now hold the real rate at zero but push inflation and the nominal rate to 10%. To keep the first-year payment under $50,000, the same household can borrow only about $470,000. The real rate has not changed. The household’s lifetime ability to pay has not changed. What has changed is the shape of the payment stream, and that shape is what banks and borrowers respond to at origination.
Evidence from consumer surveys
The authors next ask whether households behave as the theory predicts. They turn to the Michigan Survey of Consumers, which has long asked respondents both about their inflation expectations and whether they think it is a good time to buy a home. Combining the 30-year fixed mortgage rate with each respondent’s expected inflation, they construct a nominal rate and a person-specific real rate and examine which one is linked to home-buying sentiment.
The pattern is lopsided. Every one-percentage-point increase in the nominal mortgage rate is associated with an 11-percentage-point drop in the share of respondents who say it is a good time to buy. The real rate, conditional on the nominal rate, moves the other way: higher real rates are associated with slightly better home-buying sentiment, the opposite of what the standard view would predict. When respondents who say it’s a good time to buy are asked why, those who cite “low interest rates” are clearly reacting to nominal, not real, levels.
The authors add controls for unemployment and, in more recent data, for expected house price inflation, partly to address the worry that some hidden factor might be moving both home-buying sentiment and expected inflation at the same time. The coefficients barely budge.
A cross-city test
Surveys capture attitudes, not dollars. For a quantitative test, the authors turn to confidential Home Mortgage Disclosure Act data covering mortgage originations across 704 U.S. metropolitan areas from 1995 through 2023. Mortgage rates don’t vary by city, but their bite should. In places where the typical loan is large relative to income, such as Los Angeles, a rise in the nominal rate is more likely to push borrowers past the payment-to-income threshold than in a city like Houston, where loans are smaller relative to earnings.
They sort cities into three groups based on their average loan-to-income ratio and estimate how mortgage origination growth responds to changes in the nominal mortgage rate, with the real rate, house prices, the federal funds rate, and city and time fixed effects included as controls.
The pattern matches the theory. A one-percentage-point increase in the nominal mortgage rate reduces mortgage loan growth by about four to five percentage points more in the highest loan-to-income cities than in the lowest. Scaled up, the roughly four-point rise in mortgage rates between 2020 and 2022 would have cut mortgage issuance by about 16 percentage points more in high-LTI cities than in low-LTI ones. The interaction of the real rate with the high-LTI group, meanwhile, has the “wrong” sign from a standard-model perspective, echoing the survey results.
One feature of this design is that it rules out a competing explanation: money illusion, meaning borrowers simply confusing nominal and real numbers. Money illusion would not have a stronger effect in cities where payment-to-income constraints bind more tightly. A payment-to-income story predicts exactly that pattern.
Rewriting a workhorse model
In the final piece of the paper, the authors bolt a payment-to-income constraint onto an otherwise standard New Keynesian model of the macroeconomy. Households in the model consume two goods: a “cash good” paid for on the spot and a “credit good” whose purchase is capped by a payment-to-income rule that depends on the nominal rate.
Four results emerge. First, current nominal rates have real effects on output even when the real rate is held fixed, because higher nominal rates tighten the credit-good constraint. Second, because the constraint depends on today’s nominal rate rather than the full expected path of future rates, the model is less forward-looking than the textbook version. This speaks to what economists call the forward guidance puzzle, the observation that standard models imply implausibly large effects of central bank promises about future policy.
Third, higher nominal rates push demand away from the credit good toward the cash good, which reduces the marginal utility of cash-good consumption and lowers labor supply through a wealth effect. Even so, the drop in demand is large enough that output and inflation both fall when nominal rates rise. Fourth, despite these changes, the condition for a stable economy under a Taylor-type monetary rule is unchanged: the central bank still needs to raise nominal rates more than one-for-one with inflation.
When the authors extend the model to include long-term rather than one-period debt, expected future short rates re-enter the picture, but with weights that decline as the rates get further into the future. This is a feature of how long mortgage rates are built from short-rate expectations, and it is consistent with the empirical intuition that near-term monetary policy moves the cost of a 30-year mortgage more than distant ones do.
Why inflation might be more unpopular than models suggest
The authors connect their findings to a long-running puzzle: why consumers dislike inflation so intensely, even when wages keep up. Standard models struggle with the question, because they treat offsetting nominal gains and losses as a wash. The authors argue that higher inflation almost always brings higher nominal interest rates, and those higher nominal rates make it harder for households to qualify for the mortgages and car loans that finance big-ticket purchases. That effect does not show up in the Consumer Price Index, but households feel it when they try to buy a home.
A caveat worth flagging: the empirical work identifies associations across time and cities rather than clean causal shocks. The authors note that no plausible strategy separates exogenous changes in nominal rates from those in real rates, since monetary policy moves both at once. Their cross-city design addresses this by exploiting differences in how tightly payment-to-income constraints bind, rather than trying to isolate a pure nominal-rate shock at the national level.
