How Central Banks Set Rates: The Taylor Rule vs Real-World Decisions

The rule in one line

Taylor's 1993 rule, as written in a Warwick University lecture handout that reproduces it, is:

rate \= inflation \+ 2 \+ 0.5 × (inflation − 2\) \+ 0.5 × output gap

The handout describes the assumptions: the Fed "effectively followed an inflation target of 2% between 1987 and 1993, and … the long run real interest rate was also 2%." Each part has a job:

The Federal Reserve's own description uses the same coefficients. Its page on policy rules writes the rule with the longer-run neutral real rate in place of Taylor's fixed 2%, a 2 percent inflation objective, four-quarter inflation, and an output gap measured as the log difference between real GDP and potential GDP. That page was last updated in March 2018\.

• Inflation \+ 2\. If inflation is at target, the rate should at least equal inflation plus the neutral real rate, so that policy is neither stimulating nor restraining the economy.
• 0.5 × (inflation − 2). If inflation is above target, add half the overshoot on top. Together with the first term this means the nominal rate rises by 1.5 points for each 1 point of extra inflation. Because the rate rises by more than the inflation increase, the real rate goes up.
• 0.5 × output gap. The output gap is the difference between actual output and potential output, in percent. When the economy runs above potential, the rate goes up. When it runs below, the rate goes down.

A worked example

All numbers below are hypothetical and are for illustration only. Suppose inflation is 3.5%, the inflation target is 2% and the output gap is −1% (the economy is 1% below potential). We vary the assumed neutral real rate, called r*, which is the real short-term rate that would prevail with the economy at full strength and stable inflation.

(All results are our calculation.) The balanced-approach rule is the Fed's name for the same formula with the output-gap weight doubled. Two lessons show up immediately.

Small changes in inputs move the answer a lot. A 1-point change in inflation moves the prescription by 1.5 points, and a 1-point change in the output gap moves it by 0.5 points (0.5 in the Taylor rule, 1.0 in the balanced-approach rule). Moving r* from 0.5% to 2.0% moves it by 1.5 points, which is more than the whole output-gap term in this example.

The same data can support a hawkish or a dovish reading. With the economy 1% below potential, the balanced-approach rule gives a rate 0.5 points lower than the Taylor rule. If the gap were \+1%, it would give a rate 0.5 points higher (5.75% against 5.25% with r* at 0.5%). Which version a commentator picks often tracks the conclusion they already hold.

The inputs are the hard part

The Fed's page is open about this. It says that implementing a rule requires choosing an inflation measure and an activity measure, and estimating r* and whether it varies over time. It adds that it is hard to "accurately measure variables that are important determinants of the rules," and that "there is no agreed-upon 'best' model representation of the U.S. economy."

r cannot be observed. The New York Fed publishes model-based estimates (Laubach-Williams and Holston-Laubach-Williams), and states that they are "not official forecasts" of the Fed or the FOMC. The potential output behind the output gap is an estimate as well, and the Atlanta Fed's Taylor Rule Utility, which lets users choose among data sources, shows how many options exist: it draws on several r estimates, resource gaps from the Congressional Budget Office and Fed staff models, and different inflation measures. It lets users set the weight on the gap between 0 and 5 and interest-rate smoothing between 0 and 0.99.

Variants

The Fed's page lists several related rules:

To see what smoothing does, take our example with r at 0.5% and a previous policy rate of 4.0%. The inertial rule gives 0.85 × 4.0 \+ 0.15 × 3.75 \= 3.96%, essentially unchanged, even though the underlying rule prescribes 3.75%. With r at 1.5%, the rule moves up only to 4.11% against a prescription of 4.75% (our calculations). That is the point of the design. Central banks tend to move in steps rather than jump to the prescription.

• Taylor rule: output-gap weight 0.5.
• Balanced-approach rule: output-gap weight 1.0.
• Inertial rule: the rate moves only part of the way to the prescribed level each quarter. In the Fed's version, R \= 0.85 × (last quarter's rate) \+ 0.15 × (the balanced-approach prescription). The page says this damps short-term rate volatility.
• First-difference rule: prescribes a change in the rate, rather than a level. It needs no estimate of r* or potential output, but the Fed notes in a footnote that it typically creates more employment and inflation variability than the other two unless those estimates are seriously wrong.

What the Fed says it does with the rules

The Fed's page states: "Fed policymakers consult, but do not mechanically follow, policy rules." It says the FOMC regularly consults the prescriptions of several rules alongside other information, and that the rules serve as "a starting point" for judging incoming data.

Case 1: the housing boom, 2002–2006

Taylor himself argued in 2007 that policy had been too easy after the 2001 recession. Ben Bernanke's January 2010 speech to the American Economic Association summarises that finding: the federal funds rate was "below the values implied by the Taylor rule – by about 200 basis points on average over this five-year period."

Bernanke did not dispute that policy was looser than the standard rule. His reply was that the standard rule uses current inflation, while monetary policy works with a lag and should respond to forecasts. In his reading, a rule using forecast inflation prescribes a path much closer to what the Fed did. He also argued that the effect of tighter policy on mortgage costs would have been small: his simulation had the funds rate about 70 basis points higher, raising the initial adjustable-rate mortgage rate by about 0.71 percentage point and the initial payment by about $75. And he cited cross-country evidence that differences in monetary policy explain only about 5% of the variation in house price appreciation.

Taylor and Bernanke were arguing about the same rule and reached different conclusions, which shows the limits of the benchmark: the choice of inflation measure, current versus forecast inflation, and the weight on the output gap each changed the answer.

Case 2: the 2021–2022 inflation surge

In a 2023 guest post on the Econbrowser blog, economists David Papell and Ruxandra Prodan of the University of Houston used an inertial Taylor rule with an unemployment gap. They found the prescribed rate rose above the effective lower bound in June 2021, when they say the Fed fell "behind the curve." The gap between the prescribed rate and the actual rate peaked at 200 basis points in March 2022, when the Fed raised rates from near zero. It narrowed to 50 basis points by September 2022, and by December 2022 the actual rate was 25 basis points above the prescription.

This is one team's estimate using its own specification, and as the example above shows, other choices of r* or gap would shift the timing and the size. It does illustrate why rules are used: they give a reference against which to ask whether policy is behind or ahead of conditions.

How other central banks describe their decisions

The ECB describes its decisions in similar terms, without pointing to a formula. In September 2024, ECB Executive Board member Philip Lane described the ECB's approach as "a meeting-by-meeting and data-dependent approach" in which decisions rest on the inflation outlook, the dynamics of underlying inflation and the strength of monetary policy transmission. He said the ECB should "retain optionality about the speed of adjustment."

The framework behind the decisions is also reviewed periodically. On 22 August 2025 the FOMC issued a revised Statement on Longer-Run Goals and Monetary Policy Strategy, following a review that began in January 2025\. The Fed says the 2 percent longer-run inflation goal "was not a focus of the review," and that it expects to repeat such reviews roughly every five years.

What to take from this

This article is general information, not financial advice. Worked-example figures are hypothetical.

1. The rule is a benchmark, not a forecast. It tells you what a simple reaction to inflation and slack implies, not what the central bank will do.
2. Check the inputs before trusting a chart. A line labelled "Taylor rule" depends on the inflation measure, the output-gap estimate, r* and the weights. The Atlanta Fed utility shows the spread between reasonable choices.
3. A gap between the rule and the policy rate is a question, not a verdict. It can mean policy is behind or ahead of the economy, or it can mean the rule's inputs are off, or that the central bank is looking at forecasts the rule ignores.
4. Smoothing explains why rates move in steps. The inertial version captures the gradualism that central banks usually show.