Monetary Policy without Commitment
What this paper finds — and why it matters
Layer 1: Overview
Research question and motivation: Post-pandemic inflation across advanced economies rose to levels not seen since the early 1980s, reviving interest in central bank credibility. The standard quantitative macro models used to interpret this episode assume exogenous central bank reaction functions and inflation targets, which limits their usefulness. This paper instead makes monetary policy endogenous: a welfare-maximizing central bank that lacks the ability to commit re-optimizes every period. The goal is to characterize how lack of commitment shapes long-run inflation and transition dynamics, questions that prior credibility work (Barro-Gordon 1983; Rogoff 1985) could not address because it used static or log-linearized settings.
Model setup: The authors embed central bank lack of commitment into a standard fully non-linear New Keynesian model (not log-linearized around zero-inflation steady state). Monopolistically competitive firms set prices under Calvo rigidity: a random fraction 1-theta resets prices each period, the rest keep last period’s price. Wages are flexible; households choose consumption, labor, savings. The environment is deterministic with permanent unanticipated shocks. An exogenous proportional labor wedge tau (payroll tax capturing taxes, regulation, unionization) is assumed large enough (Assumption 1: tau > -1/sigma) that monopoly distortions persist. Two distortions operate: monopoly power (underproduction) and price dispersion from sticky prices (labor misallocation). The solution concept is Markov Perfect Competitive Equilibrium. Crucially, firms set prices BEFORE the central bank sets the interest rate, so the central bank takes the price distribution (hence dispersion D_t) as predetermined and optimally sets static welfare-maximizing policy: it eliminates monopoly distortions by setting the labor share to 1 (Y_t = D_t^{-1}). Equilibrium reduces to two difference equations: a forward-looking non-linear Phillips curve and a backward-looking price-dispersion law of motion, yielding a unique steady state. The analysis is conducted in a continuous-time limit for transition dynamics.
Main findings (with magnitudes and scope): (1) Long-run inflation is determined by the interaction of lack of commitment and the environment; steady-state inflation and price dispersion are strictly increasing in the labor wedge tau and strictly decreasing in the elasticity of substitution sigma (the dispersion comparative static in sigma holds for tau below a threshold tau-bar(sigma); the inflation comparative static is unambiguous). (2) Transitions to a higher-inflation steady state feature inflation OVERSHOOTING: inflation jumps on impact then gradually declines, because the central bank’s incentive to stimulate is largest early when dispersion/misallocation are low. (3) Quantitative magnitudes are large. Calibration (monthly): beta=(1.02)^{-1/12}, theta=0.86 (7-month price duration, Nakamura-Steinsson 2008), sigma=7 (Coibion et al. 2012), psi=2.5 (Chetty et al. 2011), tau=-0.1427 to target 2% annual inflation. A permanent 0.5% increase in the labor wedge raises steady-state inflation from 2% to 8.76%, with inflation overshooting to 10.11% on impact; it takes 12 months to decline within 25 basis points of the new steady state. A 0.5% decrease in sigma yields similarly large effects.
Implications: Welfare under inflation targeting strictly exceeds that under no-commitment in both shock scenarios; the welfare gain is about 6% in consumption-equivalent terms (targeting 0.981 vs no-commitment 0.922/0.921). The large magnitudes stem from a nearly vertical long-run Phillips curve (the labor share is insensitive to inflation when beta is near 1). Post-pandemic shocks (lower immigration raising the labor wedge; reduced globalization/supply-chain disruption lowering sigma) do not raise inflation on their own but do so through their interaction with central bank lack of commitment, and may make returning inflation to historic norms unlikely absent strict commitment to inflation targeting.
Layer 2: Deep Dive
What is the identification/solution strategy, and what makes the model tractable?
This is a theory paper, so ‘identification’ is the equilibrium characterization rather than econometric identification. The authors solve for Markov Perfect Competitive Equilibria of a fully non-linear (not log-linearized) New Keynesian model. Tractability comes from the timing assumption: flexible-price firms set prices BEFORE the central bank chooses the interest rate. Because the equilibrium is Markov, the central bank at date t takes the price distribution (and hence future dispersion D_{t+1} and continuation value V(D_{t+1})) as predetermined; it cannot change future welfare off the equilibrium path. So it optimally maximizes STATIC welfare conditional on current dispersion, yielding the simple first-order condition Y_t = D_t^{-1} (labor share = 1). Equilibrium then reduces to two difference equations in inflation (forward-looking Phillips curve) and dispersion (backward-looking), giving a unique steady state. A key technical innovation is an auxiliary variable delta_t (the inverse of a discounted sum of future relative prices) capturing the passthrough of real wages to current inflation holding future inflation fixed, which itself has a recursive representation and is related to the slope of the Phillips curve.
What is the core economic mechanism generating higher long-run inflation under lack of commitment?
Starting from a steady state, a permanent rise in tau (or fall in sigma) increases monopoly distortions and would, under commitment, lower the labor share while keeping inflation fixed. But a no-commitment central bank wants to undo the rise in monopoly distortions by cutting interest rates and stimulating output to push the labor share back to 1. Flexible-price firms rationally anticipate this future stimulus, higher future labor demand, and higher future real wages, so they raise prices today to offset expected future costs. Sequential price increases raise price dispersion. The economy converges to a new steady state once rising dispersion reduces aggregate productivity (labor misallocation) enough that the central bank’s marginal benefit from cutting rates vanishes. Hence both long-run dispersion and inflation are permanently higher.
Why does inflation overshoot in the transition rather than monotonically rise?
Overshooting arises from the evolution of central bank incentives as dispersion rises along the transition. Early in the transition, dispersion and labor misallocation are low, so stimulating output to boost consumption is relatively beneficial; later, once dispersion/misallocation are high, the productivity cost of stimulation is high and the benefit falls. Flexible-price firms anticipate that monetary stimulus is front-loaded, so they front-load their price increases. The result is high inflation early that declines toward the new (lower but still elevated) steady-state level. In the phase diagram (dispersion-inflation plane, holding delta fixed), the dispersion-zero locus is upward sloping and the inflation-zero locus is downward sloping; the saddle path has negative slope, so along it inflation and dispersion move in opposite directions. A labor-wedge shock shifts the inflation-zero locus up (leaving the dispersion locus unchanged); inflation jumps to the new saddle path then declines as dispersion rises.
Why are the quantitative magnitudes so large?
The steady-state labor share is relatively insensitive to inflation because the positive effect of inflation on the labor share (via overhiring sticky-price firms) is largely offset by the negative effect via forward-looking flexible-price firms that raise prices to protect against future overhiring. Standard New Keynesian calibrations use high beta and low theta, so there is a large fraction (1-theta) of flexible-price firms that raise prices substantially, putting downward pressure on the labor share. Formally, the long-run Phillips curve linking labor share mu and inflation Pi (equation 33) becomes almost vertical when beta is near 1. A nearly vertical long-run Phillips curve means small changes in tau or sigma require large changes in inflation to keep mu unchanged. Implication: any change that flattens the long-run Phillips curve would shrink the magnitudes, lower the value of commitment, and imply meaningful benefits from positive long-run inflation.
What is the central bank’s reaction function and how does it compare to a Taylor rule?
Substituting the FOC Y_t = D_t^{-1} into the Euler equation gives 1 + i_t = (1/beta) * Pi_{t+1} * Y_{t+1} * D_t. This endogenously-derived rule resembles exogenous Taylor rules: the interest rate is increasing in expected future inflation and expected future output, and it also reacts to current price dispersion. Higher dispersion reduces labor productivity via misallocation, lowering the benefit of stimulating the economy, so the central bank raises rates. Like Atkeson, Chari, and Kehoe (2010), the central bank responds to off-equilibrium increases in inflation/dispersion by raising rates enough that an individual flexible-price firm would actually want lower price increases off the equilibrium path.
How does the comparative static differ between the labor-wedge shock and the elasticity-of-substitution shock?
Both raise long-run inflation and (generally) dispersion and produce overshooting. For inflation the comparative static is unambiguous in both cases. For dispersion, the tau result is clean (Dss strictly increasing in tau), but the sigma result requires a bound: Dss is strictly decreasing in sigma only for tau < tau-bar(sigma) (where tau-bar(sigma)=infinity if sigma<=2, else 1/(sigma^2-2sigma)), because sigma also enters the dispersion law of motion and could in principle make dispersion increase with sigma when tau is large. A second difference appears in the comparison with inflation targeting: under a tau shock, an inflation-targeting central bank keeps rates fixed, output falls permanently, and dispersion is unchanged. Under a sigma shock, sigma directly affects the dispersion-inflation relationship, so even under inflation targeting steady-state dispersion would decline (greater differentiation makes relative price differences a less important source of misallocation) and rates would adjust to facilitate the transition.
What is the welfare comparison and how is welfare measured?
Welfare is expressed in consumption-equivalent terms relative to an otherwise-identical flexible-price economy: how much consumption a household would require, right after the shock, to be indifferent between the sticky-price economy (under targeting or no-commitment) and a flexible-price economy with constant consumption and implied labor. For the labor-wedge shock: welfare under targeting 0.981 vs no-commitment 0.922 (difference 0.059). For the elasticity shock: targeting 0.981 vs no-commitment 0.921 (difference 0.060). In both cases targeting strictly dominates, with gains of about 6% consumption-equivalent. The intuition: targeting reduces the misallocation cost of long-run price dispersion, while no-commitment reduces the cost of rising monopoly distortions; the dispersion costs dominate, especially because high beta makes long-run costs weigh heavily.
How does this paper relate to and differ from prior work on credibility and non-linear monetary policy?
It extends the Barro-Gordon (1983) and Rogoff (1985) credibility tradition, which used static or linearized settings that cannot speak to long-run inflation or transition dynamics. It differs from Markovian linearized approaches (e.g., Halac and Yared 2022) which feature no transition dynamics and significantly OVERESTIMATE the effect of permanent shocks on long-run inflation (because linearization underestimates the welfare cost of rising dispersion). It departs from fiscal-commitment models (Alvarez-Kehoe-Neumeyer 2004; Aguiar et al. 2015) and from Davila-Schaab (2023, which uses quadratic adjustment costs and thus has no price dispersion) by emphasizing the Calvo dispersion cost and its dynamic feedback on the inflation-output tradeoff. Relative to the discretionary-multiplicity literature (Albanesi-Chari-Christiano 2003; King-Wolman 2004; Zandweghe-Wolman 2019), this model obtains a UNIQUE equilibrium and provides an analytical (not numerical) characterization of the steady state and transition. It also contributes a novel recursive representation of the non-linear Phillips curve via the auxiliary variable delta_t.
What are the transition dynamics of the macro variables in the calibrated exercise?
Following the permanent labor-wedge increase: inflation jumps up from 2% and gradually declines toward its higher steady state (overshooting). The nominal interest rate jumps up and continues rising throughout the transition (the higher steady-state nominal rate reflects the Fisherian effect present in the non-linear model). The real interest rate jumps DOWN initially (the central bank stimulates to weather the shock) then gradually returns to its original level. Output falls gradually as price dispersion and labor misallocation increase. Nominal wage inflation jumps up with price inflation but stays below it, converging from below; this gap underpins a permanent long-run decline in the real wage.
What are the policy implications and their scope conditions?
Permanent changes in the global economy (e.g., lower immigration shifting labor toward more regulated/higher-wedge sources; slower globalization or supply-chain disruptions raising domestic firms’ market power, i.e., lower sigma) can raise long-run inflation, but only through their interaction with central bank lack of commitment, not on their own. The post-pandemic inflation spike, and its overshooting, can be partly understood as the private sector rationally anticipating accommodative policy. Scope condition: this holds as long as the central bank operates with FULL DISCRETION; a strict commitment to inflation targeting would prevent it. There can therefore be significant benefits to institutions that enhance commitment. A caveat from the model’s own logic: if structural changes flatten the long-run Phillips curve, magnitudes shrink, the value of commitment falls, and there are real benefits to positive long-run inflation (so targeting too low an inflation rate would be costly).
What are the main caveats and directions for future research the authors flag?
The model is deterministic with permanent shocks and abstracts from monetary-fiscal interactions by assuming lump-sum taxes and Ricardian equivalence (debt is payoff-irrelevant, set to zero). It focuses on the stable steady state, setting aside equilibrium implementation and off-equilibrium inflation stability. The discretionary policy (labor share = 1) is invariant to the price-setting model, so the approach extends to menu-cost or rational-inattention models. Future work: relax Ricardian equivalence to study interactions between central bank and fiscal lack of commitment (facilitated by the framework not assuming a long-run debt level since it is not linearized), and examine off-equilibrium inflation stability.