The Reaction Function: How Policy Is Actually Set

The rule, and why it is a benchmark rather than a policy

The prescription has three parts: a neutral real rate, which is the rate that neither restrains nor stimulates once inflation is at target; inflation itself, which has to be added because the policy rate is nominal, so a rate held constant while inflation falls is automatically tightening; and adjustment terms for the two gaps, which give the rule its stabilising character. Each term has a line of argument behind it. The neutral rate is unobservable and estimated with wide uncertainty. The coefficient of half on inflation is a convention, and a rule with a coefficient below one is unstable — a rise in inflation lowers the real rate and invites more inflation. The output-gap term is what makes policy act before inflation actually moves. Applied to the current economy, the rule prescribes about 4.05% against an actual midpoint of 3.875%. That agreement is not a verdict about policy; it is a statement that, on these input values, the committee is roughly where a simple rule would put it. The analytical value comes from moving the inputs. Let inflation fall to target with the output gap unchanged and the prescription falls to 2.70%. Let the output gap turn to −1% and the prescription is 3.35% even with inflation 0.9 points above target — which is the entire trade-off the committee faces, written as one line of arithmetic. Raise the neutral rate to 2.0%, as the post-pandemic debate about the long-run neutral rate suggests it may be, and the same economy prescribes 5.55%. That last comparison is the lesson. Half a point of disagreement about a variable nobody can observe moves the prescription by a point and a half, which is more than most tightening cycles achieve in a year. So the market does not forecast the economy and apply a rule; it forecasts the committee, using the data as evidence about which reaction function is in force. That is why the same unemployment print can move the expected path in opposite directions in different regimes, and why a change in the committee's stated framework is a bigger event than a data surprise of the same size. One economy, four prescriptions — Core inflation 2.9%, output gap +0.4%, neutral 0.5%: 4.05% — about where policy is · Inflation back at target, other inputs unchanged: 2.70% — the easing a benign forecast implies ← · Output gap −1.0%, inflation still 2.9%: 3.35% — a weak economy argues for easing despite above-target inflation · Neutral real rate 2.0%: 5.55% — an unobservable half-point moves the answer by 1.5 points No central bank follows a rule mechanically, and quoting one as though it does misreads the institution. Committees weigh financial conditions, credibility and the risk of being wrong in public — which is why a rule is best used to describe what a reasonable committee would do with your forecast, not to predict what it will do with theirs.

From the policy rate to the assets you hold

The policy rate is a single short-term nominal rate, and it reaches portfolios through four channels. The first is the discount rate: the risk-free rate anchors every valuation, so a higher expected path lowers the present value of long-dated cash flows, and it lowers them most for the assets with the longest duration — long bonds and long-duration equities. The second is credit: higher policy rates raise the cost of new borrowing and the burden of floating-rate debt, which shows up first in the weakest balance sheets and widens spreads. The third is the currency, because capital moves toward the higher real return and a stronger currency imports disinflation and hurts exporters. The fourth is wealth and expectations: higher rates reduce asset values, which reduces spending, and they also reduce inflation expectations, which is the channel the committee needs. Those four channels have different lags — credit and currency respond within quarters, while the effect on inflation through demand takes a year or more — which is precisely why policy is described as acting with a lag and why tightening cycles end before inflation reaches target. For an investor, the useful translation is that the expected path is a statement about the economy wearing the costume of a rate. When the path steepens because growth expectations improved, equities can rise with it; when it steepens because inflation expectations rose, both legs of a balanced portfolio fall at once. That distinction — is the move about real growth or about inflation — is the one to make before deciding what a repricing means for the book, and it is the same distinction the next lesson applies to real and nominal returns. • Discount rate: higher expected path lowers the value of long-duration cash flows most. • Credit: the weakest balance sheets reprice first, and spreads widen. • Currency: capital moves to the higher real return, which imports disinflation. • Expectations and wealth: the channel the committee needs, and the slowest. A rate path is a forecast of the economy in disguise. Steepening on better growth and steepening on higher inflation look identical on the chart and mean opposite things for a portfolio.

The inputs are themselves estimates

A policy rule looks mechanical, and the arithmetic is: a target rate from the neutral rate, the inflation gap and the output gap. But two of the four inputs cannot be observed. The neutral rate — the level consistent with stable inflation at full employment — is inferred from models and moves as productivity and demographics change; it has been estimated anywhere from near zero to several percent depending on the method and the decade. The output gap, the distance from potential, is computed from an estimate of potential that is itself revised for years. That means a rule does not deliver a number, it delivers a **range**. Two reasonable estimates of the neutral rate and the output gap can put the prescription two full percentage points apart, which is the difference between easing and holding. The professional use of a rule reflects this: it is a benchmark that makes a policy stance legible — “the actual rate is well below the rule, so policy is accommodative” — rather than a forecast of the next decision. Debating a rule to a tenth of a point is debating the error bars. The practical takeaway for markets is that the rule is one of several inputs the committee weighs, alongside financial conditions, the labour market and incoming data. The parts of the reaction function that surprise markets are rarely the rule’s arithmetic; they are a change in how the committee is weighting its unobservables — a shift in the estimated neutral rate, or a new emphasis on the labour market over inflation. If a model’s key inputs are unobservable and have wide error bands, its output is a range. Report the range, or the number will mislead.

The rulebook has more than one page

There is no single policy rule, and the differences between the leading candidates are not academic — they say different things about how fast a committee should move and how much of the past it should carry. The benchmark is the original specification in this lesson: a neutral real rate, plus inflation, plus half of each gap. Three variants matter in practice. The **inertial** version adds a term for the previous policy rate, which is a way of writing down the fact that central banks move gradually and smooth through noise; its effect is a rule that responds less to any single data point, which is exactly what a committee that has been burned by one-month surprises wants. The **balanced-approach** variant gives the output gap a larger coefficient, which argues for keeping policy easier when unemployment is high even if inflation is above target — the trade-off stated as a number rather than a preference. And the **first-difference** version drops the level of the rate altogether and responds only to changes in inflation and unemployment, which implies that the level inherited from the past should carry no weight at all; it is the rule of someone who thinks the previous path was simply wrong. The framework the committee operates under sits above the rule, and it changes less often and matters more. In 2020 the Federal Reserve adopted a flexible form of average inflation targeting: after a period in which inflation ran below two percent, policy would aim to have inflation run moderately above two percent for some time, rather than treating two percent as a ceiling to be defended at every meeting. The change had a specific consequence for the rule: in a world where the neutral rate is low, the zero lower bound was doing most of the tightening, and the framework was a way of promising to make up lost ground so that expectations would not drift down. It also means that reading the framework from the statement is a distinct exercise from reading the rate from the gaps — the same inflation print implies different policy depending on which averaging window the committee has committed to, and the commitment itself is re-examined on a five-year cycle, which makes the release of a framework review a first-order market event rather than a committee curiosity. The practical point of the menu is that a rule reading is only meaningful when the rule is named. “The economy is below the rule” has no content without the specification behind it: a first-difference rule can say tighten while the benchmark says hold, on identical data. So the discipline is the one the rest of this subject teaches in other places — write down the specification, the inputs and the vintage of the data, and then treat the disagreement between rules as the measure of how much the decision depends on an assumption nobody can observe. When the rules agree, the policy path is nearly mechanical; when they disagree by a full point, the market is not forecasting the economy, it is forecasting which page of the rulebook the committee is reading from. Four specifications, one economy — Benchmark Taylor: Neutral real rate + inflation + ½ of each gap — the reference case · Inertial: Adds the previous policy rate, so any single print moves the prescription less ← · Balanced approach: A larger output-gap weight: hold easier when unemployment is high ← · First difference: Responds only to changes, implying the inherited level carries no weight The framework adds a fifth axis underneath all four: whether the target is treated as a ceiling or as an average to be achieved over time. That choice is what decided how the committee read the inflation overshoot after 2021, and it is why the statement’s language is examined for framework signals rather than only for rate signals.

Sources

• The rule and its original statementTaylor (1993), “Discretion versus Policy Rules in Practice”
• Rule-based policy analysis and the neutral rateFederal Reserve education materials; FOMC statements and projections, September 2026
• The market’s expected path and the term structureStandard term-structure literature; futures-implied policy path methodology