John B. Taylor | Macro Paper Warehouse

Discretion versus policy rules in practice

This 1993 Carnegie-Rochester Conference paper by John Taylor introduces what became known as the “Taylor rule” — a simple guideline for setting the federal funds rate, r = p + 0.5y + 0.5(p - 2) + 2, where p is inflation over the previous four quarters and y is the percent deviation of real GDP from a 2.2%-per-year trend — and argues for treating monetary policy rules as broad, “systematic and credible” approaches to policymaking rather than as mechanical algebraic formulas to be followed literally. Drawing on multicountry rational-expectations model comparisons (his own and others’), Taylor reports that interest-rate rules responding to inflation and output outperform rules targeting the money supply or a fixed exchange rate, and that a flexible exchange-rate regime dominates a fixed one on both output-stability and price-stability grounds for the G-7 countries he examined. The paper’s hypothetical policy rule, despite its deliberately round-number coefficients, is shown to track actual Federal Reserve behavior over 1987-1992 remarkably closely, with a notable exception in late 1987 when the Fed eased following the stock market crash. Taylor then works through two real episodes — the 1990 oil-price shock following Iraq’s invasion of Kuwait, and the sharp 1990 rise in long-term interest rates coinciding with German unification — to argue that departures from a rule’s literal prescription can themselves be principled and rule-consistent once policymakers correctly diagnose whether a disturbance is temporary (the oil shock, confirmed by futures prices showing only a modest expected persistent rise) or driven by real rather than inflationary forces (the German rate rise, attributable to a unification-driven investment-demand shift rather than inflation expectations).

Expectations, Open Market Operations, and Changes in the Federal Funds Rate

This paper develops a simple daily model of the U.S. federal funds market to explain how a Federal Open Market Committee (FOMC) announcement of a new target for the federal funds rate can move the actual rate almost the full distance to the new target on the same day, even though the New York Fed’s Trading Desk typically conducts no open market operation designed to bring that change about until the following day. The model has two parts: a “Trading Desk reaction function” (drawing on the supply-side literature) in which the Desk adjusts the supply of Fed balances the day after the effective funds rate deviates from target, and a demand for Fed balances (building on recent microeconomic work by Furfine 2000a and by Guthrie and Wright 2000) in which banks’ demand today depends in part on their rational expectation of tomorrow’s funds rate. Because traders know the Desk will act tomorrow if today’s rate has not yet converged to target, the anticipation of that future action shifts today’s demand for balances and moves today’s rate immediately – a mechanism Taylor labels, following Guthrie and Wright’s New Zealand terminology, an “open mouth operation.” Simulating the calibrated model, Taylor shows a 50-basis-point target increase can move the effective rate by roughly 42 basis points on the announcement day alone, with the residual gap closing geometrically over the following days as the Desk’s actual (lagged) reserve adjustments catch up; the same model, applied to a demand shock rather than a target change, reproduces the observed speed with which funds-rate deviations from target revert to zero. Using daily 1998-2000 U.S. data, Taylor also documents that the volatility of funds-rate deviations from target declined and their day-to-day persistence rose over this period, particularly around actual target changes, and that the target rate Granger-causes the funds rate far more consistently than the reverse – patterns broadly consistent with the model’s predictions, though Taylor cautions the model’s timing is sharper than what the data actually show and that some anticipated target changes visible in fed funds futures markets produce little detectable movement in the funds rate itself.

The Monetary Transmission Mechanism: An Empirical Framework

This 1995 Journal of Economic Perspectives paper by John B. Taylor lays out an empirical framework for the monetary transmission mechanism and asks what it implies for the choice between fixed and flexible exchange rates and for the design of a monetary policy rule. Rather than identifying shocks in a VAR, Taylor synthesizes his own multicountry structural model (Taylor 1993a) – a rational-expectations model with staggered wage and price setting, estimated for the US, Canada, Germany, France, Japan, Italy, and the UK – together with corroborating reduced-form evidence from Romer and Romer (1994). The framework centers on a five-link causal circle: a change in the short-term interest rate moves the exchange rate through uncovered interest rate parity, and that real exchange rate change affects net exports and GDP; the short rate also moves the long-term rate through an expectations model of the term structure, and that real long rate affects consumption and investment (including all three components of fixed investment – business equipment, business structures, and residential – which are significantly negatively related to the real interest rate in the US, and negatively related to the real interest rate in every G-7 country); GDP and inflation then feed back into the short rate through the policy reaction function, closing the loop. Taylor argues financial-market prices – the short rate, the exchange rate, and the long rate – are the right objects of analysis because money demand for M1 and M2 “seems to have shifted substantially in recent years” and credit-flow measures are “at least as unreliable,” while uncovered interest rate parity is said to describe the “larger and longer swings” in interest rate differentials and the dollar reasonably well even though high-frequency deviations remain unexplained. Comparing model elasticities estimated over an early-1970s-to-mid-1980s sample against an early-1970s-to-mid-1990s sample, Taylor reports that the real GDP response to a 3 percent unanticipated permanent increase in the target price level (a proxy for an easing) fell in all three countries examined in detail – the US, Germany, and Japan – with the decline larger in the US than in the other two; within the US the interest elasticity of investment fell while that of consumption rose, though “there is no general pattern” across the G-7. Simulations of exchange rate arrangements among the US, Japan, and Europe show flexible rates producing lower variability of both real GDP and inflation than fixed rates, and Taylor notes that his own 1993 interest-rate rule – weighting the inflation deviation and output gap equally at 0.5 – “has turned out to describe recent Federal Reserve policy very accurately.” He also states that bond-supply effects on the term structure, while sometimes detected, are “too small to explain the errors in the term structure equations on a systematic basis.” Throughout, the underlying elasticities and simulation results are drawn from the companion structural model (Taylor 1993a) rather than estimated afresh in this paper, and Taylor flags the framework’s own weak points: the short rate is “only one of many factors” affecting the exchange rate and long rate, uncovered interest rate parity’s high-frequency failures remain unexplained, and the choice of a single long-term real rate to represent the cost of capital is an acknowledged simplification.

The Power of Open-Mouth Policies

When a central bank announces a policy it will not implement for some time, households and firms respond straight away. This paper measures how much of a policy’s effect arrives in that anticipation window. The obstacle is technical: an announced, dated, one-off policy change is not a recurrent draw from a stationary process, so it produces a nonstationary solution — a different decision rule in every period — that conventional methods, which construct a single time-invariant decision rule, cannot represent. The authors develop a perturbation-based extended function path (EFP) method to build that sequence of time-dependent decision rules, and apply it to a scaled-down replica of the Bank of Canada’s ToTEM projection model. Across five experiments they find the anticipation effects are large for two of them and modest for the other three: a gradual rise in the inflation target from 2% to 3% raises output by about 0.2% at its peak when implemented immediately and by about 0.3% when announced a year ahead, and forward guidance about lifting off from the effective lower bound raises the peak output response by about 70% when the return to the Taylor rule is postponed from four quarters to eight — while switching to a more aggressive Taylor rule, to price-level targeting or to average inflation targeting produces only minor anticipatory effects in an economy not otherwise hit by shocks. Comparing their solution with a Markov news-shock treatment of the same experiment, they find the Markov version “significantly overstates the importance of a given anticipated event”, with peak anticipation effects in investment five times larger, because a unit-root news process implies the effects persist forever. (The full text read for this summary is the authors’ manuscript of November 16, 2024, which carries the same abstract, model and five experiments as the published article; magnitudes may have moved in revision.)