The Neutral Rate: What the Long Run Pays

What the neutral rate is, and why it cannot be measured

The neutral rate is the real interest rate at which the economy operates at potential with inflation stable — neither pushing demand up nor holding it back. It is the anchor a policy maker compares the actual real rate against, and the whole framework of this subject depends on it: the reaction function uses it directly, the curve prices it as the long-run level the path converges to, and the debt arithmetic turns on it. Its nominal version is simply the neutral real rate plus the inflation target, which is what makes it the long-run anchor of the entire yield curve rather than one input among many. The measurement problem is fundamental rather than technical. It cannot be observed because it is a counterfactual: the interest rate that would keep the economy at potential is a property of a world nobody can visit. Instead it is inferred, either from the behaviour of inflation and output through a model of how the economy responds to rates — the approach behind the published estimates, which infer r* from the joint behaviour of inflation, output and the policy rate — or from market prices, using the gap between indexed bonds and nominal ones as a proxy with its own distortions. Both approaches produce a number with a confidence band wide enough to matter, both are revised as data arrive, and both disagreed with each other through the twenty-tens by more than the range of policy decisions taken in a year. Two properties of the estimates are worth internalising. First, they are only visible with hindsight, so the estimate available in real time is the least reliable version of it; a central bank acting on a mis-measured neutral rate will tighten too little for too long, which is the error the framework in the nineteen-seventies institutionalised. Second, the uncertainty is asymmetric in its consequences: if r* is higher than assumed, policy is looser than intended and inflation drifts up, while if it is lower, policy is tighter than intended and the economy runs below potential — and the first error is the one that costs the anchor. The same economy at three estimates of r* — r* = 0.2% (low end of credible estimates): prescription 3.75% · r* = 0.5% (central estimate): prescription 4.05%, against an actual midpoint of 3.875% · r* = 1.5% (high end): prescription 5.05% ← · Spread across that range: 1.30 percentage points — more than a year of ordinary tightening · Long-run nominal neutral at r* = 0.5%: 2.50%, against a 10-year yield of 4.20% The last row is where this lesson meets the term-premium one. If the long-run nominal anchor is 2.5% and the ten-year yields 4.20%, the gap is either a higher expected policy path than the anchor implies or a term premium — and separating the two is exactly the exercise that split could not do on its own.

What moves it, and what a lower r* does to everything else

The drivers are the forces that set the balance between the supply of saving and the demand for capital, because the neutral rate is the price at which that balance clears. On the saving side: demographics, since an ageing population saves more and invests less; inequality, since a high-income household saves a larger share of its income; and the global demand for safe assets, which is not a saving decision so much as a willingness to accept a lower return for a claim that is guaranteed — and which suppresses the observable safe rate without doing much for the return on capital. On the investment side: productivity growth, since faster growth raises the return on capital and therefore the rate that clears; the capital intensity of the economy; and the size of the investment demand that comes from a structural programme such as an energy transition, a reshoring cycle or a large and permanent fiscal deficit. Estimates through the twenty-tens fell on the saving side of this ledger and produced the low-rate equilibrium everyone learned to expect; the debate since then has been about whether the investment side has shifted, and a credible answer today is a range rather than a point. The consequences of a low r* run through every other lesson in the subject. It lowers the long-run anchor and therefore the whole curve, which matters because every asset’s required return is benchmarked to it — a lower r* is a permanent re-rating of long-duration cash flows, which is why the level of rates and the level of multiples are related over long periods. It turns the debt arithmetic: sustainability is about r minus g rather than the level of debt, so a low r* gives a government more fiscal room than the debt ratio alone suggests, and a rising r* takes it away (MR11 for the arithmetic). And it makes the effective lower bound bind more often, which is not a small operational detail: with a neutral real rate near zero, an ordinary recession can push the required rate below the floor, so policy has to be delivered through the balance sheet and through guidance rather than the rate (MR15), and the asymmetry means the average inflation outcome has a downward bias unless the framework compensates for it. A low r* is therefore a low-margin environment, and low-margin environments are where policy errors persist. • Drivers sit on the saving side (demographics, inequality, safe-asset demand) and the investment side (growth, capital intensity, structural investment). • The estimates are only reliable with hindsight, and they are revised as data arrive. • A lower r* lowers the long-run anchor and re-rates long-duration cash flows. • It changes the debt arithmetic through r minus g rather than through the debt level. • And it makes the effective lower bound bind more often, so the balance sheet becomes a policy instrument. Treat any single estimate of the neutral rate as one draw from a wide distribution. The honest use of r* is not to compute a prescription to a decimal place — it is to check whether the direction of policy is right and how much of the answer depends on an assumption, which is exactly what moving the estimate across its range shows.

The estimates, and the range that is the story

The neutral rate is unobservable, and that has not stopped a small industry from estimating it. Knowing how the estimates are produced matters less than knowing what their dispersion means, because the spread across estimates is usually wider than the policy disagreements that cite them — and the spread *is* the finding. The standard approaches are model-based. One family, associated with Laubach and Williams, treats the neutral rate as the level of the real short rate consistent with output at potential and stable inflation, and estimates it jointly with the economy’s trend growth and the other unobservables using a small macro model. A later version extends the same idea to a panel of economies, which is useful because it borrows information across countries and produces estimates for economies whose own data is too short to fit alone. The estimates are computed from revised data, move when the data moves, and carry confidence intervals that are wide enough that the point estimate should be regarded as the middle of a band. Then there are the cheaper readings, each with a specific weakness. The **market-implied** approach subtracts expected inflation and a term premium from a forward rate: simple, observable, and contaminated by whatever the premium is doing — which is the problem the previous read examined. A **survey** of professional forecasters asks directly and is transparent, at the cost of depending on the forecasters. And a **growth-based** proxy takes long-run potential growth as an anchor for the neutral rate on the argument that capital’s return and the economy’s growth are linked: crude, defensible, and useful mainly as a sanity check on the model output. The dispersion across these methods is not small — enough to change the sign of the answer to “is policy restrictive”, which is exactly the question the number is used to settle. That is the honest conclusion: the neutral rate is a framework for organising a discussion about policy stance rather than a measurement that settles one. A central bank acting on a point estimate is acting on a number whose uncertainty exceeds the gap it is trying to close. Two practical consequences follow. For policy, the response is to move gradually and to watch the data rather than the estimate, because a wrong estimate of the neutral rate shows up as a persistent inflation or slack surprise rather than as an immediate error. For an investor, the response is to hold a range rather than a number and to ask which side of the range matters more: a higher neutral rate raises every discount rate and pressures long-duration assets, while a lower one supports them, so the position you would regret being wrong about is the one to size against. • Model estimates carry confidence intervals wider than the policy questions asked of them. • Market-implied and survey readings each inherit a specific weakness. • The spread across methods can change the sign of “is policy restrictive”. • Hold a range, and size against the side of it you would regret being wrong about. A practical test for any commentary citing the neutral rate: ask for the range. Where the answer is a single number, the sentence is describing a model output with more confidence than the model supports — and the confidence, rather than the level, is usually the thing being communicated.

Sources

• Estimates of the natural rate of interestHolston, Laubach & Williams, “Measuring the Natural Rate of Interest: International Trends and Determinants”; Laubach & Williams (2003)
• Secular stagnation and the balance between saving and investmentLawrence Summers, “U.S. Economic Prospects: Secular Stagnation, Hysteresis, and the Zero Lower Bound” (2014)
• Safe-asset demand and the convenience yieldKrishnamurthy & Vissing-Jorgensen on the supply and demand for Treasury securities; standard safe-asset literature
• r minus g and debt sustainabilityMR11 on the debt arithmetic in this curriculum; standard fiscal-sustainability literature