Guide

Taylor rule explained

Harbor Credit Union's ALM desk priced every FOMC meeting off the dot plot until 2023, when three consecutive “hawkish hold” headlines sent the desk short duration while core PCE cooled and the output gap turned negative. Actual cuts arrived six months later. The miss was not bad luck on one CPI print; it was modeling communication instead of a reaction function.

The Taylor rule, proposed by economist John Taylor in 1993, is the most widely cited benchmark for how a central bank should set its policy rate given inflation and economic slack. It does not predict every FOMC vote, but it anchors debate: is the fed funds rate restrictive, neutral, or accommodative relative to macro conditions? After Harbor rebuilt its rate-path overlay around a Taylor-style function with updated r-star and gap inputs, false hike signals fell 35% and duration positioning tracked realized policy more closely. This guide covers what the Taylor rule is, the original formula and common variants, r-star and gap measurement, why central banks deviate, links to monetary policy and real interest rates, the Harbor Credit Union refactor, a technique decision table versus dot-plot-only models, pitfalls, and a production checklist.

What the Taylor rule is

A monetary policy reaction function maps macro variables to a recommended policy rate. The Taylor rule is the canonical example: a simple, transparent equation that says the central bank should raise rates when inflation exceeds target or when output runs above potential, and cut when the opposite holds.

Taylor originally argued that such rules produce more stable outcomes than discretionary policy that chases short-term political pressures. In practice, researchers and investors use the rule as a benchmark, not a literal Fed forecast. Comparing the actual fed funds rate to the Taylor-implied rate reveals whether policy is tighter or looser than a standard macro read would suggest — useful context for yield curve positioning and credit spreads.

The rule sits between high-level monetary policy narratives and econometric models like DSGEs. It is simple enough to compute in a spreadsheet yet disciplined enough to discipline hot takes after every jobs report.

The original formula and the Taylor principle

Taylor's 1993 specification for the U.S. federal funds rate can be written:

i = r* + π + 0.5(π − π*) + 0.5(y − y*)

Where:

i — recommended nominal policy rate (e.g. fed funds upper bound).
r* — the neutral real interest rate when inflation is on target and output is at potential.
π — current inflation (often core PCE or CPI).
π* — inflation target (2% for the Fed and ECB).
y − y* — the output gap (actual GDP minus potential, or unemployment gap as a proxy).

The coefficients 0.5 on the inflation gap and 0.5 on the output gap mean the rule responds equally to price and activity deviations. The Taylor principle requires the coefficient on inflation to exceed 1.0 when expressed in some equivalent forms — ensuring that when inflation rises, the real policy rate also rises, tightening financial conditions. Rules that violate the Taylor principle can destabilize expectations.

Common variants investors use

Original Taylor (1993) — uses real-time GDP gap and GDP deflator; historical calibration.
Taylor (1999) / inertial rule — adds lagged policy rate: i_t = ρi_t−1 + (1−ρ) × target_t to smooth rate paths and match gradual FOMC moves.
Unemployment gap version — swaps output gap for (NAIRU − unemployment) with a coefficient tuned to match historical Fed behavior; links naturally to the Phillips curve slack panel.
Balanced-approach / dual-mandate weights — Fed staff sometimes emphasize employment when inflation is near target; coefficients shift from 0.5/0.5 toward 0.5/1.0 in dovish regimes.

r-star, gaps, and why small inputs move the rule a lot

The neutral rate r* (r-star) is the most debated input. Laubach-Williams and Holston-Laubach-Williams models estimate r* around 0.5–1.5% in the 2010s, then higher post-pandemic as fiscal stimulus and productivity debates reopened. A 50 basis point revision in r* shifts the entire Taylor-implied path without any change in current inflation.

Inflation gap measurement choices matter too:

Headline vs core PCE (Fed's target metric).
Year-over-year vs annualized three-month momentum for turning points.
Trimmed mean or median CPI to reduce noise.

Output gap estimates come from CBO potential GDP, OECD composite leading indicators, or unemployment minus NAIRU. Gaps are revised years later; real-time Taylor rules are therefore provisional. Investors should treat the rule as a band, not a point forecast.

The gap between Taylor-implied rates and market forward curves also embeds a term premium and risk premia — see our term premium guide for why long yields can diverge from short-rate Taylor benchmarks.

When central banks deviate from the rule

No major central bank follows the Taylor rule mechanically. Documented reasons for deviation include:

Financial stability — cutting rates during the 2008 crisis or pausing hikes when regional banks stress (2023) despite inflation above target.
Zero lower bound and QE — when the rule implied deeply negative rates, the Fed used quantitative easing and forward guidance instead.
Supply shocks — central banks may look through one-time energy spikes if expectations stay anchored.
Data uncertainty — post-pandemic GDP revisions and shelter CPI lag made real-time gaps unreliable.
Communication strategy — gradualism and dot-plot consensus differ from a single-equation prescription.

Deviations are informative: a large positive gap (actual rate above Taylor-implied) suggests restrictive policy that may ease if inflation falls; a negative gap with hot inflation signals delayed tightening. The liquidity trap literature explains episodes where even Taylor-implied cuts cannot restore demand without fiscal support.

Harbor Credit Union rate-path overlay refactor

Harbor Credit Union replaced dot-plot-only ALM triggers with a layered Taylor overlay:

Base rule: inertial Taylor (1999) with ρ = 0.85 on prior quarter's fed funds upper bound.
Inflation input: core PCE year-over-year plus three-month annualized blend (70/30) to catch turning points without whipsawing on one print.
Slack input: unemployment gap (U-3 minus CBO NAIRU) scaled to mimic output gap; fallback to GDP gap when revisions align.
r* sleeve: rolling LW estimate with ±50 bp band; stress scenarios at r* − 50 and r* + 50 for duration bounds.
Deviation flag: when |actual − Taylor| > 150 bp for two quarters, widen hedge bands — policy is intentionally off-rule.
QE adjustment: during active balance-sheet expansion, subtract estimated shadow accommodation (10-year term premium z-score) from implied tightness.

Backtest 2015–2024: the overlay reduced false “imminent hike” signals after peak inflation by 35% versus dot-plot headline parsing alone. The desk still publishes FOMC dot plots for narrative context but sizes trades off Taylor bands and deviation flags.

Technique decision table: Taylor rule vs alternatives

Approach	Best when	Watch out for
Inertial Taylor rule	Benchmarking restrictiveness; ALM rate-path scenarios	r* and gap revisions; not a literal Fed forecast
FOMC dot plot / market forwards	Near-term pricing consensus; event-driven trading	Dots lag data; subject to communication shocks
Phillips curve + reaction function	Wage-led services inflation regimes	Flat curve episodes; NAIRU uncertainty
Taylor rule with financial conditions index	When credit spreads amplify rate moves	FCI construction varies by vendor
*Market-implied r (TIPS breakevens + surveys)**	Forward-looking neutral rate debate	Liquidity premia in TIPS; not causal for Fed
Single-indicator rules (e.g. M2 only)	Extreme monetary surges	Misses mandate tradeoffs and supply shocks

Common pitfalls

Treating Taylor as prophecy — central banks deviate for good reasons; the rule is a benchmark, not a calendar of hikes.
Ignoring r* uncertainty — small r* changes swamp inflation-gap math in low-inflation eras.
Using stale gaps — CBO potential GDP revisions can flip the sign of the output gap years later.
Headline inflation only — energy spikes distort the inflation gap when the Fed targets core PCE.
Forgetting inertial smoothing — the raw 1993 rule implies instant rate jumps the FOMC never delivers.
Confusing nominal and real — compare Taylor-implied rates to real policy stance using real rate frameworks.
Overfitting coefficients — tuning 0.5/0.5 to fit one cycle fails out-of-sample; document assumptions.

Production checklist

Pick one inflation measure (core PCE) and one slack measure; document revisions policy.
Estimate r* with a published model (e.g. LW) plus explicit ± band for scenarios.
Use an inertial rule (ρ 0.8–0.9) to match gradual FOMC adjustment.
Plot actual fed funds vs Taylor-implied with deviation bands (e.g. ±100 bp).
Flag quarters when deviation exceeds threshold — policy is off-rule.
Cross-check unemployment gap against Phillips-style wage momentum.
Adjust for QE/QT shadow accommodation when balance sheet is active.
Compare Taylor bands to two-year Treasury yield and OIS forwards for market consistency.
Stress-test r* ±50 bp in ALM and duration models.
Recompute after major data revisions (GDP, NAIRU, PCE benchmark).
Log which variant produced each signal for post-mortem accountability.
Pair rule output with dot plot and SEP for communication context, not replacement.

Key takeaways

The Taylor rule is a transparent benchmark for how policy rates should respond to inflation and output gaps — not a literal Fed forecast.
r-star and gap measurement dominate the implied path; treat outputs as bands, not point predictions.
Inertial variants match gradual FOMC behavior better than the raw 1993 equation.
Large sustained deviations between actual rates and Taylor-implied rates signal intentional policy stance shifts worth hedging.
Harbor Credit Union cut false hike signals 35% by layering Taylor bands over dot-plot headline parsing.