Real interest rates, the returns on lending after stripping out inflation, don’t just wander randomly. They drift in long, slow arcs that can last decades. They climbed through the late 1970s and early 1980s, then declined for nearly forty years before turning back up. Economists have offered plenty of explanations, from aging populations to slower productivity growth to a global appetite for safe assets. But a working paper suggests that a large piece of the puzzle has been sitting in plain sight: the public’s stubborn, memory-driven beliefs about inflation, and central bankers’ ongoing efforts to push back against them.
The NBER working paper by Stefan Nagel of the University of Chicago Booth School of Business argues that when households and firms form their long-run inflation expectations by learning from what they’ve personally lived through, central banks can’t simply talk those expectations back into line. Instead, policymakers have to keep interest rates persistently high (or low) for years to move realized inflation enough to eventually shift beliefs. That long tug-of-war, Nagel argues, leaves a distinctive fingerprint on real rates across the entire term structure.
The idea behind “learning from experience”
The starting point is a framework Nagel developed with Ulrike Malmendier in earlier work. In it, people don’t form inflation expectations by reading Federal Reserve speeches or studying economic models. They form them by paying attention to the inflation rates they’ve actually seen during their own lifetimes, with recent experiences weighted a bit more heavily than distant ones.
This produces a specific pattern: older Americans, who remember the double-digit inflation of the 1970s, tend to expect higher inflation than younger Americans. And because personal memory fades slowly, the average person’s long-run inflation expectation drifts only gradually. Nagel estimates that a fresh quarterly inflation surprise nudges the average household’s long-run belief by only about 1.8%.
That slowness matters. If people’s beliefs about long-run inflation adjust only inch by inch, then a central bank facing inflation expectations that have drifted above its target has a problem. Announcements and policy statements can’t reset those beliefs. Only sustained periods of actual low inflation will do it, and generating sustained low inflation requires holding interest rates high for a long stretch.
A model of the tug-of-war
Nagel builds this intuition into a version of the standard New Keynesian model that macroeconomists use to study monetary policy. In his version, households and firms have subjective beliefs shaped by experience rather than perfect foresight, and they have finite planning horizons, meaning they don’t try to forecast the economy indefinitely far into the future.
Within this model, the central bank behaves rationally: it understands that today’s inflation expectations will linger, so when those expectations sit above its target, it pushes nominal interest rates up more than one-for-one to slowly grind them back down. From the private sector’s point of view, though, this looks different. Households and firms don’t realize the central bank is temporarily leaning against their beliefs. They interpret the high real rates as a permanent shift in what economists call the natural rate of interest, the neutral rate consistent with a stable economy.
Because private-sector agents view the shift as permanent, long-term real interest rates move about as much as short-term real rates. The whole term structure shifts together. And because those planning horizons are finite, these big shifts in long-term rates don’t produce wild swings in current output.
Testing the prediction against the data
To see whether this story holds up, Nagel builds a quarterly time series of experience-based long-run inflation expectations for the United States, running from 1961 through 2025. He uses inflation data going back to the late 19th century to construct each birth cohort’s lifetime inflation memory, then averages across cohorts aged 25 to 74 to get an aggregate expectation. He deliberately uses the same weighting parameter estimated from survey data on actual inflation expectations, rather than picking one that would maximize the fit to interest rates.
He then compares this expectations series against real interest rates measured several ways: ex-post realized real rates (nominal rates minus what inflation actually turned out to be), ex-ante rates using household inflation expectations from the Michigan Survey of Consumers, and ex-ante rates using professional forecasters’ expectations. He looks at short-term rates, five-year rates, and seven-year forward rates.
The relationships are strong. A one percentage point rise in experience-based long-run inflation expectations is linked to roughly a one percentage point increase in ex-ante short-term real rates, and to a similarly sized increase in real rates seven years out on the forward curve. Because the underlying series are highly persistent, Nagel uses statistical techniques designed for that situation, including a bounds test that formally rejects the possibility that the observed correlations are spurious byproducts of shared trends.
Adding this experience-based component to standard estimates of the natural rate of interest helps explain patterns those estimates miss on their own. Real rates were lower than the estimated natural rate in the 1960s and early 1970s, higher in the 1980s and 1990s, and lower again more recently. Much of the decades-long slide in real rates from the early 1980s until about 2020, according to Nagel’s analysis, is associated with monetary policy gradually working down the inflation expectations inherited from the 1970s.
The same pattern shows up abroad
Nagel then repeats the exercise for Germany, the United Kingdom, and Japan, using each country’s own inflation history but the same weighting rule estimated from U.S. survey data. In all three countries, ex-post realized real rates line up with experience-based long-run inflation expectations in a way that mirrors the U.S. results.
Japan is a particularly interesting case because its real rates fell to near zero in the mid-1990s, well ahead of other advanced economies. That distinct path also tracks the country’s experience-based inflation expectations, which came down earlier. Nagel handles two thorny data problems along the way: Germany’s Weimar-era hyperinflation and Japan’s postwar hyperinflation. He treats these episodes as regime changes that reset beliefs and winsorizes quarterly inflation at 10% to prevent extreme observations from swamping the calculation.
Connecting to bond market puzzles
The framework also touches on several oddities in bond market behavior that other models struggle to explain. Long-term bond yields move more closely with short-term rates than standard theory predicts. They react strongly to macroeconomic announcements, even at horizons of 10 to 15 years out. And most of the excess returns earned on long-term Treasury bonds over the past several decades occurred on days when major macroeconomic data was released.
Nagel argues these patterns fit the experience-based story. If investors are constantly updating their long-run inflation views based on incoming inflation data, then inflation surprises will move long-term yields. And if inflation expectations drifted persistently downward from the early 1980s through 2020, investors would have been repeatedly surprised by lower-than-expected inflation, generating a sequence of unexpectedly high returns on long-duration bonds. Under this interpretation, those returns don’t represent compensation for risk. They reflect forecast errors that ran in the same direction for a very long time.
What the argument does and doesn’t claim
Nagel is explicit that his mechanism is meant to complement, not replace, other explanations for secular real-rate movements. Demographics, productivity trends, and safe-asset demand can all still play a role in determining the natural rate of interest. What he argues is that on top of those forces, there’s a separate source of persistent real-rate variation that arises from the interaction between memory-based expectations and monetary policy responses.
The framework also implies something uncomfortable for central bankers: because expectations shaped by lived experience don’t respond directly to communication, a central bank that has lost credibility on inflation can’t restore it with words alone. It has to produce sustained low inflation, which requires holding real rates elevated for years. Nagel notes that this stands in tension with the rational-expectations view in which credible policy announcements can move expectations instantly, though he leaves room for exceptions in cases like the ends of hyperinflations that Thomas Sargent studied, where dramatic changes in fiscal regimes appear to have reset beliefs quickly.
