Oil Prices, Monetary Policy and Inflation Surges

Layer 1: Overview

Gagliardone and Gertler ask why the US inflation surge that began in mid-2021 was both sudden and persistent, and whether a simple structural model can account for it without targeting inflation in estimation. The paper’s central claim is that the surge was driven primarily by the combination of large oil price shocks and accommodative (“easy”) monetary policy by the Federal Reserve, with oil complementarities and real wage rigidity as the key amplification mechanisms. Secondary factors — demand shocks and labor-market tightening — matter but do not drive the surge on their own.\n\nThe model is a New Keynesian framework with three non-standard features relative to the Blanchard-Gali (2007) benchmark: (1) oil enters both household utility and firm production as a complement rather than a substitute (elasticities of substitution estimated at ψ = 0.02 for households and ε = 0.37 for firms, both well below unity); (2) a Mortensen-Pissarides search-and-matching labor market that makes unemployment endogenous and allows shocks to matching efficiency; and (3) real wage rigidity parameterized by γ, estimated at 0.697, meaning actual wages adjust only about one-third as much as Nash bargaining wages would.\n\nEstimation uses simulated method of moments, matching model impulse responses to two sets of SVAR impulse responses identified via high-frequency external instruments: oil-price surprises around OPEC announcement dates (following Känzig 2021) and monetary-policy surprises around FOMC dates (following Gertler-Karadi 2015, extended by Bauer-Swanson 2022). The SVAR sample runs 1973:01–2019:12, with 2020–2022 reserved as an out-of-sample validation window. The model is then taken to the 2010–2022 period for a historical shock decomposition, targeting unemployment, real oil price inflation, the Federal Funds rate, and labor-market tightness; headline and core PCE inflation are left entirely untargeted and used as the key test of model fit.\n\nMain quantitative findings: the estimated elasticity of substitution between oil and labor in production is ε = 0.37 (s.e. 0.16) and between oil and consumption goods for households ψ = 0.02 (s.e. 0.34), both significantly below unity and confirming strong complementarity. Real wage rigidity γ = 0.697 (s.e. 0.145): actual wages move roughly one-third as far as Nash wages. The Calvo price parameter λ = 0.945 implies an average price duration of approximately six quarters at monthly frequency, and habit persistence h = 0.914.\n\nIn the structural VAR, a monetary tightening of 15 basis points reduces GDP by about 10 basis points (peak after ~10 months) and raises unemployment by roughly 0.5 percentage points; a 6 percent increase in the real oil price reduces GDP 20–30 basis points and raises the core PCE price level about 20 basis points. Complementarities matter quantitatively: at the estimated parameters, the peak GDP drop following an oil shock is 0.13 percent versus only 0.04 percent under Cobb-Douglas (no complementarity), and the core PCE inflation response is more than double in the benchmark. The decline in the marginal product of labor accounts for more than half the increase in marginal cost during the 2021 surge.\n\nIn the historical decomposition (2010–2022), oil shocks and easy monetary policy shocks jointly account for the bulk of the 2021–22 inflation surge; labor-market matching shocks contribute little to either unemployment variation or inflation; demand shocks dominate unemployment variation but are not the primary inflation driver in the surge. The model also explains the 2014–2019 low-inflation/low-unemployment puzzle: declining oil prices and tight money shocks kept inflation down despite a tight labor market, the mirror image of 2021–22. Baseline forecasts (as of spring 2023) under a Taylor rule with coefficient 2 project headline and core PCE declining to roughly 3 percent in about one year then converging slowly to 2 percent, with unemployment rising to approximately 5 percent (its steady state) and overshooting by about half a percentage point. A more aggressive tightening (funds rate held at 4.6 percent through September 2023) reduces inflation by about half a percentage point faster but raises unemployment by an additional persistent 1 percentage point.

Layer 2: Deep Dive

What is the identification strategy for the oil and monetary policy shocks, and what are the main threats?

Both shocks are identified as external instruments in an SVAR. The oil shock uses daily surprises in oil futures prices on days of OPEC meetings (Känzig 2021): the surprise is the change in the log oil futures price between the day before the meeting and the close on the announcement day. The money shock uses surprises in the first principal component of the first four quarterly Eurodollar futures in a 30-minute window around FOMC announcements and non-FOMC Fed communication dates (Gertler-Karadi 2015, extended by Bauer-Swanson 2022). The key identifying assumption is relevance and exogeneity: each surprise must be correlated with the structural shock of interest but uncorrelated with the other structural shocks. The primary threat addressed is endogeneity between oil prices and monetary policy: oil price movements prior to FOMC meetings predict the monetary policy surprise (coefficient 0.073, s.e. 0.038), plausibly because the Fed responds systematically to energy prices. The authors regress money surprises on the monthly log change in oil spot prices and use residuals as the cleaned monetary instrument. Without this purging, the SVAR counterfactually predicts a surprise tightening raises oil prices. The authors also drop the Lehman Brothers date from the sample because confounds from the financial collapse would distort the monetary impulse response. A secondary threat is the use of a daily (rather than intraday) window for oil surprises, justified by evidence that oil markets react more slowly to OPEC announcements than financial markets react to FOMC meetings.

How does strong complementarity between oil and labor amplify the inflation response, and how is this mechanism isolated empirically?

With a CES production function where ε < 1, firms cannot easily substitute away from oil when its price rises. The marginal product of labor declines sharply because each worker needs roughly the same amount of oil to be productive, raising marginal cost of output for any given wage. The Phillips curve then transmits this cost-push increase to inflation. The authors show analytically that the sensitivity of the marginal product of labor to the ratio of oil to labor is proportional to 1/ε: as ε falls, the oil shock’s impact on marginal cost and hence inflation rises sharply. This is isolated by comparing the benchmark model against a Cobb-Douglas version (ε = 1, ψ = 1): peak GDP decline is 0.13 percent with complementarities versus 0.04 percent without; the unemployment response is large and persistent only with complementarities; and the core PCE inflation response is more than double in the benchmark. The historical decomposition further shows that the decline in the marginal product of labor accounts for more than half the increase in marginal cost during the 2021 surge.

What role does real wage rigidity play, and what is the resulting inflation-unemployment trade-off?

Real wage rigidity introduces a cost-push term into the Phillips curve. Without rigidity (γ = 0), the Nash bargaining wage absorbs the oil shock, and the central bank can achieve both price stability and efficient employment simultaneously. With γ = 0.697, actual wages fall by only about one-third as much as Nash wages after an oil shock. The gap between Nash and actual wages enters the Phillips curve as a cost-push term Δt. If the central bank tries to stabilize prices, it must contract demand enough to push the efficient component of marginal cost negative, forcing output and unemployment well below the flexible-price equilibrium — in the model, pursuing price stability after an oil shock causes output and unemployment to deviate from the flexible-price benchmark by more than double over the first 8–10 months. This trade-off rationalizes partial monetary accommodation and is quantitatively important for matching the historical behavior of inflation in 2021–22.

How does the historical shock decomposition work, and what are its key identifying assumptions?

The authors use the estimated DSGE model with the Kalman smoother to perform a historical shock decomposition over 2010–2022. They estimate persistence and standard deviations of four shocks (demand εbt, monetary policy εrt, oil εst, and matching efficiency εΦt) using Bayesian methods, targeting four observable series: unemployment, real oil price inflation, the Federal Funds rate, and labor-market tightness from JOLTS. Nominal variables — headline PCE, core PCE, nominal wage growth, real product wage growth — are entirely untargeted and serve as out-of-sample validation. One important wrinkle is that the spot oil price contains high-frequency speculative volatility that does not pass through to the prices households and firms face. The authors filter this by assuming nominal oil price inflation equals PCE energy inflation plus an i.i.d. speculation shock, so that only the persistent component enters real allocations. The posterior mean of the speculation shock standard deviation (σm = 0.239) is substantially larger than that of the persistent oil shock (σo = 0.042), confirming the filter’s importance.

What sub-sample variation is documented, and what explains it?

The model resolves three sub-sample puzzles. First, the 2014–2019 period had low unemployment but persistently low inflation — the model attributes this to declining oil prices and tight monetary policy shocks that offset demand pressures and kept marginal cost subdued. Second, the 2010–2012 period had rising oil prices but also low inflation — attributable to a large negative demand shock from the Great Recession lingering, which depressed marginal cost sufficiently to offset the oil price effect. Third, the high labor-market tightness of 2022 is shown to be largely an endogenous response to easy monetary policy and oil shocks rather than an autonomous labor supply shock. The matching shock does not materially contribute to either unemployment variation or inflation over the sample.

What robustness checks are reported?

(1) Taylor rule coefficient: calibrating ϕπ to 1.5 instead of 2 adds roughly 0.5 percentage points to PCE inflation at the peak of the 2022 surge due to money shocks but does not change qualitative conclusions. (2) Matching shock persistence: results are robust to calibrating persistence to 0.9 or 0.95 instead of the estimated 0.548, confirming that the matching shock’s minimal contribution to inflation is not an artifact of low persistence. (3) Unemployment demeaning: using 6 percent instead of 5 percent does not change results. (4) Oil price speculation filter: removing the filter has only minor quantitative effect because anomalous spike-and-reversal days are few. (5) Monetary policy shock orthogonalization: without purging oil-price predictability from the money surprise, the SVAR counterfactually predicts tightening raises oil prices, confirming the necessity of the adjustment.

How does this paper relate to and differ from Blanchard and Gali (2007)?

The paper descends most directly from Blanchard-Gali (2007), which also features oil in a New Keynesian model with real wage rigidity. Key differences: (i) Gagliardone-Gertler make oil a complement rather than a substitute or Cobb-Douglas input in both utility and production, which they argue is necessary to match quantitatively the observed impact of oil shocks on inflation; (ii) they incorporate a Mortensen-Pissarides search-and-matching labor market with endogenous unemployment, enabling labor-market tightness to function as a separate inflation driver; (iii) they estimate the model formally by matching SVAR impulse responses to externally identified shocks rather than calibrating; and (iv) they apply the model specifically to explaining the 2021–22 inflation surge. The real wage rigidity mechanism is retained from Blanchard-Gali as a central feature.

How does this paper relate to the broader literature on the 2021–22 inflation surge?

The paper explicitly positions itself against work emphasizing supply chain disruptions and goods-sector reallocation (Guerrieri et al. 2021, Di Giovanni et al. 2022, Ferrante et al. 2023) as the main drivers of 2021 inflation. The authors accept that supply chains mattered in 2021 but argue they moderated by end of 2021 while inflation persisted through 2022, so their framework targets the more durable sources. Papers closer in spirit emphasize monetary policy (Ball et al. 2022, Amiti et al. 2022, Benigno-Eggertsson 2023, Pflueger 2023), but Gagliardone-Gertler differ by using a structural DSGE model estimated to identified shocks and by giving oil shocks a prominent co-equal role alongside monetary accommodation. Lorenzoni and Werning (2023) share the emphasis on production complementarities and wage rigidity.

What are the policy implications and their scope conditions?

The primary policy implication is that the 2021–22 inflation surge was jointly caused by oil shocks and monetary accommodation, and unwinding it involves a short-run cost in real activity due to the inflation-unemployment trade-off generated by real wage rigidity. The baseline forecast is slow convergence to 2 percent inflation with a quasi soft landing: headline and core PCE reaching roughly 3 percent in about one year then declining slowly, and unemployment rising to 5 percent steady state and overshooting by about half a percentage point. A more aggressive tightening (funds rate at 4.6 percent through September 2023) brings inflation to 2 percent faster by about half a percentage point by June 2023 but at the cost of an additional persistent unemployment increase of about 1 percentage point. Scope conditions: (i) results depend critically on long-run inflation expectations remaining anchored at 2 percent — if expectations drift to 3 percent, the disinflation task becomes harder; (ii) the model abstracts from supply chain disruptions, downward nominal wage rigidity, and open-economy channels; (iii) the quantitative conclusions rest on estimated complementarities that carry large standard errors, especially for household oil complementarity ψ.

What is the role of labor-market tightness as an inflation driver in this framework?

Labor-market tightness (θt = vt/ut) raises marginal cost through two channels: it increases net hiring costs (a tighter market requires more vacancies to fill a given number of positions, raising the per-hire cost) and it raises the Nash bargaining wage (because unemployment becomes less painful, improving workers’ outside option). In the historical decomposition, however, the matching efficiency shock — the exogenous source of tightness variation — contributes negligibly to both unemployment variation and inflation over the 2010–2022 sample. The high tightness of 2022 is shown to be largely an endogenous response to easy monetary policy and oil shocks rather than an autonomous labor-supply disruption. This finding challenges the narrative that autonomous labor-market tightening was a primary independent cause of the inflation surge.

Key Concepts

Oil complementarity (ε, ψ): In the paper’s CES framework, oil is a complement when the elasticity of substitution with labor in production (ε) or with consumption goods for households (ψ) is below unity. A value below unity means that when oil becomes scarce, the marginal productivity of labor (or marginal utility of other consumption) falls more than proportionally, amplifying the macroeconomic impact of oil price shocks. Estimated values of ε = 0.37 and ψ = 0.02 imply strong complementarity in both sectors.

Real wage rigidity (γ): A parameter ∈ [0,1] measuring how sticky the actual real wage is relative to the Nash bargaining wage. With γ = 0.697, the actual wage moves only about one-third as far as the Nash wage in response to a shock (wqt = (w°qt)^{1−γ}(wq)^γ). This is adopted as a reduced-form mechanism — not derived from deeper frictions — that generates realistic unemployment volatility and introduces a short-run inflation-unemployment trade-off absent from fully flexible-wage models.

Cost-push term (Δt): The component of inflation in the Phillips curve that arises purely from the gap between actual wages and Nash bargaining wages when real wage rigidity is present. Equals −κγ times the deviation of the Nash wage from steady state. It is the mechanism through which oil supply shocks create an inflation-unemployment trade-off: even if the central bank stabilizes the efficient component of marginal cost, the cost-push term generates inflation, and offsetting it requires contracting demand below the efficient level.

Impulse-response matching estimation: The paper’s estimation procedure: simulated method of moments minimizes the weighted squared distance between model-implied impulse responses and SVAR-estimated impulse responses to externally identified oil and monetary shocks. Precision weights from the SVAR IRF confidence bands determine which moments receive more weight. Confidence intervals for structural parameters are obtained via the delta method. This approach ensures the model can simultaneously explain the dynamics following both supply (oil) and demand (monetary) disturbances.

Easy monetary policy shock: A negative realization of the monetary policy shock εrt in the Taylor rule, representing the actual Federal Funds rate falling below what the estimated Taylor rule coefficient on inflation would prescribe. In the historical decomposition, such shocks from roughly mid-2020 onward are attributed substantial responsibility for low unemployment and upward pressure on inflation in 2021–22, distinct from endogenous policy responses to demand or oil shocks.

Speculation shock (εmt): An i.i.d. component of nominal oil price changes that is not reflected in the PCE energy price index and therefore does not pass through to real allocations in the model. Introduced to prevent high-frequency gyrations in spot oil prices (attributed to financial-market speculation) from generating counterfactually large macroeconomic swings. Its estimated standard deviation (posterior mean 0.239) is substantially larger than that of the persistent structural oil shock (0.042).

Historical shock decomposition (untargeted nominal variables): The primary empirical test of the model: after estimating shocks from four targeted real/financial series (unemployment, real oil price inflation, Federal Funds rate, labor-market tightness), the model constructs predicted paths and shock contributions for headline PCE inflation, core PCE inflation, nominal wage growth, and real product wage growth — none of which were targeted in identification. Agreement between model predictions and data for these untargeted nominal variables is the main evidence that the model correctly identifies the sources of the inflation surge.