What Is R-star? The Neutral Real Interest Rate Explained

Economists use r-star as shorthand for a neutral or natural real interest rate. In a common monetary-policy definition, it is the short-term real rate that would prevail when output is near its sustainable level and inflation is stable. A real rate adjusts a nominal interest rate for inflation. The benchmark is called neutral because, under the definition and assumptions of a model, it is neither adding pressure to demand nor restraining it.

The phrase “natural rate” can have narrower theoretical meanings, and researchers do not all use exactly the same horizon or equilibrium conditions. Some estimates aim at a longer-run rate that moves slowly; others try to capture shorter-run forces. The New York Fed's Laubach-Williams and Holston-Laubach-Williams estimates, for example, model r-star as a latent real short-term rate associated with output near potential and stable inflation. The model's assumptions are part of the result, not a footnote that can be ignored. See the [New York Fed's r-star overview]({source:nyFedRStarEstimates}).

R-star is therefore best read as a conditional reference point. It helps organize a question about policy stance, but it does not say what interest rate is morally “correct,” guarantee an economic outcome, or reveal what a central bank will do next. Before comparing two r-star numbers, check whether they refer to the same country, time horizon, definition, model, and data vintage.

The federal funds rate is a nominal overnight interest rate. The Federal Open Market Committee sets a target range for it and the Federal Reserve uses its policy tools to guide market rates toward that range. R-star is an estimated real benchmark. The FOMC can choose the nominal target; it cannot set the economy's underlying neutral rate directly.

To compare a nominal policy rate with a real r-star estimate, first put them on a comparable real basis. A rough ex-ante real policy rate is the nominal policy rate minus expected inflation over a matching horizon. It is an estimate because expected inflation is not directly observed in a single definitive series. The comparison also needs compatible frequencies and definitions: an overnight real rate should not be casually compared with a distant long-run estimate as if they were identical objects.

If the estimated real policy rate is above the chosen r-star estimate, analysts often describe policy as restrictive relative to that benchmark. If it is below, they may call policy accommodative. This shorthand is conditional. The gap is uncertain, the transmission of policy takes time, and other financial conditions can move in a different direction. The labels do not tell a borrower what rate they will receive or a household how much its mortgage payment will change.

Suppose a purely hypothetical policy-rate midpoint is 4.0%, and expected inflation over the relevant short horizon is 2.5%. A rough expected real policy rate is 4.0% − 2.5% = 1.5%. If one model estimates r-star at 0.5%, the gap is 1.5% − 0.5% = positive 1.0 percentage point. Under that model, policy would be above its neutral benchmark.

Now keep the same hypothetical real policy rate but assume a different defensible model estimates r-star at 2.0%. The gap becomes 1.5% − 2.0% = negative 0.5 percentage point. The qualitative reading changes. This is not a trick in the arithmetic: it shows that a stance label inherits uncertainty from both expected inflation and the estimated benchmark.

These made-up values are only for demonstrating the subtraction. They are not current U.S. rates, a forecast, or an estimate of the FOMC's view. Real analyses should state the nominal rate used, the expected-inflation measure and horizon, the r-star estimate and its horizon, and the date of the underlying data. If those details are missing, “above neutral” can sound more precise than the evidence allows.

The neutral rate can change as households' desired saving and firms' demand for investment evolve. Faster expected productivity or potential-output growth can raise the return on productive investment and may put upward pressure on equilibrium real rates. A greater desire to save, weaker investment demand, or stronger demand for safe assets can work in the other direction. Demographics, public borrowing, risk preferences, and conditions in the global economy can also matter, with effects that vary by model and period.

There is no single stable coefficient that converts any one factor into a fixed change in r-star. For example, population aging might increase saving in one setting, while longer retirements could change the timing of dissaving in another. Productivity can support investment demand, but the effect on saving and investment rates depends on expectations, financing, and the distribution of gains. These mechanisms are reasons to study the benchmark; they are not a calculator for a precise future value.

Some estimates distinguish a slowly moving long-run component from temporary forces that shift the short-run neutral rate. A temporary change in demand or financial stress may alter the rate consistent with balanced conditions for a while without proving that the economy's long-run destination changed. Always match a statement about “r-star” to the intended horizon.

No exchange posts a price labeled “the neutral real rate.” Economists can observe nominal interest rates, inflation measures, output, and other data, but the equilibrium rate is an inferred variable. It is defined by a counterfactual condition: what real short rate would be consistent with the model's assumptions about stable inflation and sustainable output?

One family of models uses relationships among real GDP, inflation, and the federal funds rate to infer unobserved trends and the neutral rate. State-space approaches estimate observed series alongside hidden quantities such as trend growth, potential output, the output gap, and r-star. The model must distinguish a temporary shock from a lasting shift in potential growth or the equilibrium rate, so a recent point estimate does not show all the assumptions behind that distinction. The New York Fed publishes Laubach-Williams and Holston-Laubach-Williams model estimates, accompanying data, and code. It also states that these model estimates are not official forecasts of the New York Fed, its president, the Federal Reserve System, or the FOMC.

Another analyst may use a different structural model, a statistical filter, a survey, or market information. These methods do not necessarily target the same concept, and survey or market measures can include risk premia or other influences in addition to an expected real rate. Different results do not automatically mean that one analyst made an arithmetic mistake: models can transform the data differently, let trends move at different speeds, impose different relationships between inflation and output, or define the equilibrium condition differently. Compare the definition and method alongside the reported value. A Summary of Economic Projections participant's longer-run federal funds rate projection, an FOMC decision, a staff model estimate, and a New York Fed r-star series are issued for different purposes and are not interchangeable.

An estimate is not simply “the rate the Fed thinks is neutral.” It is the output of a named method using chosen data and assumptions. The model can be useful even though it is uncertain: it forces the analyst to spell out a mechanism and compare observed developments with a consistent benchmark. It becomes misleading when the model's conditional estimate is reported as a directly observed price or as a precise official target.

<!-- learn:illustration -->

Economic data are revised, and a model's reading of the latest period can change when a new observation arrives. Estimating a hidden trend at the present edge of a data series is especially difficult because the analyst has no future observations to help distinguish a lasting shift from a temporary movement. Later data can make the estimate for the past look different from the estimate that was available at the time.

This is called a real-time estimation problem. A real-time estimate uses only information that had been published by the date being studied. A later or smoothed estimate can use additional observations from after that date. The two answer different questions: what policymakers could have inferred then, and what a model estimates now with more history. The San Francisco Fed's discussion of [estimating the neutral real rate in real time]({source:sfFedRStarRealTime}) describes how one-sided estimates can differ from estimates that use later data.

For example, if a statistical agency revises real GDP growth for several years in the past, the revised output path can alter both actual output and the model's estimate of potential output. The output gap and inferred trend growth may shift, which can change a model's historical r-star path as well. A revision published today does not prove that policymakers had the revised information at the time. Evaluating a past decision requires separating the real-time information set from a model's current reconstruction.

The endpoint of the sample is another difficulty. A date in the middle of a data series has observations before and after it, but the latest quarter has no future data yet. Once another quarter arrives, the model may revise its estimate of recent hidden trends by more than estimates for earlier years. State whether a policy analysis uses the estimate that was available in real time or a later estimate reconstructed with additional history.

Model uncertainty is separate from data revisions. Two specifications can process the same data and still produce different paths because they impose different relationships, priors, or definitions. A published estimate should be read with its model name, publication date, data vintage, and uncertainty range when one is provided. A single decimal point does not remove those limits.

R-star is not the 10-year Treasury yield, a mortgage rate, or the interest rate on a savings account. Long-term market yields reflect expected future short rates, inflation compensation, term premiums, liquidity, and other factors. A short-term neutral real-rate estimate is a different object. It may inform how one interprets policy, but it cannot substitute for analyzing a particular bond or loan.

It is also not the same as the Federal Reserve's 2% longer-run inflation objective. The objective is a nominal inflation goal; r-star is a real interest-rate concept. Adding expected inflation to a long-run real neutral estimate can provide a rough nominal neutral benchmark, but that calculation depends on the inflation expectation and definitions used. The Federal Reserve's [Summary of Economic Projections FAQ]({source:federalReserveSummaryEconomicProjections}) explains that the SEP includes individual participants' longer-run federal funds rate projections; those are not a survey of one model's r-star value and do not commit the FOMC to a particular path.

Finally, an estimated policy gap is not a complete summary of financial conditions. Credit spreads, exchange rates, asset prices, lending standards, and fiscal developments can affect demand. A central bank also weighs its goals, forecasts, uncertainty, risks, and the effects of policy with a lag. The San Francisco Fed's analysis of [the asymmetric costs of misperceiving r-star]({source:sfFedMisperceivingRStar2021}) shows why mistaken estimates can have unequal consequences; it does not turn any one published estimate into a complete policy verdict.

Start by identifying the exact estimate: who produced it, which model or survey it uses, what definition and horizon it targets, when it was released, and which data vintage it used. Then identify the real policy rate being compared with it. For a nominal policy rate, state how expected inflation was measured and match its horizon to the rate. If the source reports a range or uncertainty interval, keep that range visible instead of quoting only the midpoint.

Next, treat “restrictive” or “accommodative” as a conclusion conditional on the chosen benchmark, not as an observable label. Ask whether the comparison is about today's short-run stance or a longer-run equilibrium. Consider whether revisions, a different model, or a change in inflation expectations could alter the gap. These checks are more informative than debating a single tenth of a percentage point without naming the assumptions.

R-star connects to the real-versus-nominal interest rate, the Taylor rule, and estimates of the output gap. Those concepts work together in policy analysis, but each represents a different quantity. A useful reading keeps those quantities separate and treats every neutral-rate estimate as a model-based guide rather than a promise about future policy or market returns.

This guide describes U.S. monetary-policy concepts. It does not estimate a current r-star, forecast an FOMC decision, or recommend a financial product.