Managing an Energy Shock: Fiscal and Monetary Policy

The paper argues that neither of the two existing strands of the energy-shock literature can capture the direct link from high energy prices to depressed aggregate demand that policymakers were worried about during the 2021-2023 European energy crisis (Sec. 1). Papers studying the supply side (e.g., Baqaee and Farhi 2019, 2022; Bachmann et al. 2022) find only modest GDP effects from realistic substitution elasticities but abstract from nominal rigidities entirely, so they have no aggregate demand channel. Papers that do have an aggregate demand channel – New Keynesian models with oil, following Blanchard and Galí (2007a) or Bodenstein, Erceg and Guerrieri (2011) – generate recessions only through the endogenous monetary tightening that follows the shock’s inflationary impulse, because their complete-markets or permanent-income households have essentially zero marginal propensity to consume out of an energy-price-induced income loss. The paper’s own contribution is to show that adding realistic household heterogeneity restores a direct real-income-to-spending channel, independent of any monetary policy response, addressing concerns raised explicitly by ECB chief economist Philip Lane about the “adverse income and wealth effects of rising energy import prices on aggregate demand.”

In the complete-markets representative-agent model, aggregate consumption is completely unaffected by an energy price shock (dC = 0), so the entire output response comes from expenditure switching toward domestic goods, and it is always positive (Prop. 1, eq. 28-30). This follows from combining the Backus-Smith condition with the real UIP condition under a constant-real-rate monetary rule: since the real exchange rate and hence relative consumption stay pinned at steady state, output rises by an amount proportional to the ratio of the energy share of consumption to the domestic-good share, scaled by a weighted-average elasticity of substitution χ. The paper stresses this does not mean RA models can never predict a bust after an energy shock – when they do (as in Bodenstein et al. 2011), it is because of monetary tightening in response to the shock, not the shock’s direct effect on demand; in the language of the standard three-equation New Keynesian model, an energy shock in an RA framework is best interpreted as a cost-push shock paired with a positive aggregate demand shock.

With heterogeneous agents and incomplete markets, the matrix of intertemporal MPCs M is no longer zero, opening up a real income channel that works against expenditure switching, and there is an exact neutrality threshold at χ = 1 where the HA and RA models coincide (Prop. 2-3, eq. 31-33). At χ = 1 – which here corresponds exactly to every underlying elasticity of substitution equaling 1, the Cole and Obstfeld (1991) unitary-elasticity case – the real income channel and the multiplier exactly cancel, leaving output driven purely by expenditure switching, just as in the RA model. Below χ = 1, expenditure switching weakens, the real income channel dominates, and the HA output response falls below the RA response; at the paper’s realistic calibrated value of χ = 0.3 (and illustratively for values around χ ≈ 0.5 and below), the shock causes a sizable contraction rather than the RA model’s expansion, and by χ = 0.1 the contraction is large.

No – adding a real-wage stabilization motive to the union’s wage-setting problem (following Blanchard and Galí 2007b) generates a genuine wage-price spiral, in which both nominal wages and prices rise together, but this is entirely self-defeating and never actually protects real wages, whose decline is fully pinned down by the shock itself independent of the nominal wage Phillips curve (Sec. 3.3, eq. 34, Fig. 5). Mechanically, unions pushing up nominal wages raises domestic goods prices, which depreciates the exchange rate, which raises import prices in turn, so that the entire CPI basket ends up rising by exactly enough to match the nominal wage increases. Without the real-wage stabilization motive, by contrast, the model instead predicts an initial price-level jump followed by a sustained decline in both wages and prices, as depressed hours and consumption raise households’ willingness to work at any given wage.

No to both: energy entering the production function rather than the consumption basket produces an isomorphic GDP response under an intuitive re-scaling of parameters (Prop. 4), and simulating the actual, large, nonlinear 2022 energy shock as a full perfect-foresight (“MIT shock”) path rather than a small first-order shock does not materially change the qualitative or quantitative conclusions (Sec. 3.4). The production-side equivalence requires only that the effective household spending shares on home, foreign, and energy goods line up across the two model versions; the paper also shows that allowing the country to produce some of its own energy (an “energy endowment”) mitigates but does not eliminate the contraction at low substitution elasticities, since redistribution toward lower-MPC households still occurs, and that an incomplete-markets representative-agent model (RA-IM), while technically featuring positive iMPCs, has such small ones that it “struggles to generate substantial contractionary effects without very long-lived shocks.”

A single small energy-importing country’s interest rate has essentially no power over the world price of energy – the paper’s back-of-the-envelope UIP calculation implies that a one-percentage-point rate increase sustained for a year appreciates the currency by only about one percent – so unilateral tightening buys little disinflation at a real cost in lost output, whereas coordinated tightening across all energy importers materially lowers world energy prices and therefore delivers much more disinflation for the same output cost (Sec. 4, 6, Fig. 19). In the paper’s simulated coordinated tightening, world energy prices fall by around 35 percentage points on impact, and the resulting reduction in CPI inflation is “more than twice as much” as under uncoordinated tightening, even though pass-through of world prices into domestic consumer prices is slow. This creates what the paper frames as a free-rider problem: since each individual central bank’s tightening has a small direct effect on its own inflation but a positive externality on other importers via lower world energy demand, uncoordinated central banks may rationally choose looser policy than would be collectively optimal.

No – imported energy inflation is harder for monetary policy to stabilize than domestic inflation of the same initial magnitude, especially once downward nominal wage rigidity is present, because monetary tightening’s main channel for reducing CPI inflation from an energy shock is contracting output to slow wage growth, a channel that downward wage rigidity partly shuts off (Sec. 4.3, Fig. 14). For a domestically-generated shock of the same size, by contrast, wages are themselves pulling up the CPI, so downward nominal wage rigidity does not bind in the same way, and monetary policy is correspondingly more effective at achieving a given amount of disinflation for a smaller output cost.

All three deficit-financed fiscal instruments the paper studies – energy subsidies, targeted transfers indexed to a household’s counterfactual energy consumption, and untargeted lump-sum transfers – substantially limit the output and consumption fallout from the shock, but they differ sharply in their inflation consequences: transfers raise CPI inflation (mainly via a stronger wage-price response to higher aggregate demand), while subsidies restrain it (Sec. 5.2, Fig. 15-16). Because subsidies directly lower the real energy price households actually pay, they defuse the wage-price spiral at its source, reducing union pressure for nominal wage increases; transfers, in contrast, support consumption by stimulating aggregate demand from households with high MPCs (per the logic of Farhi and Werning 2016 and Auclert, Rognlie and Straub 2023), which pushes wage and price inflation up rather than down. All three policies also reduce the rise in cross-sectional consumption inequality caused by the shock (Fig. 17), though somewhat less effectively if the government lacks fiscal space and must run something closer to a balanced budget.

Because subsidies blunt every subsidizing country’s incentive to substitute away from energy, so when many energy importers adopt them simultaneously, world energy demand becomes almost price-inelastic and the world energy price spikes sharply, eroding much of the very insulation the subsidies were meant to provide while still leaving governments to finance the associated fiscal cost (Sec. 6, Fig. 20). In the paper’s coordinated-world simulation, this endogenous rise in world energy prices pushes CPI inflation back up “to similar levels as without energy subsidies,” and the recession actually worsens relative to the uncoordinated case, because governments must raise labor income taxes further to finance the now much larger subsidy bill – making energy subsidies the fiscal policy with the largest negative cross-country externality of the three the paper studies, even though it is also the most effective at shielding an individual country acting alone.

Yes – using Romer and Romer (2004) monetary policy shocks and a Jordà (2005) local-projection specification on U.S. data (1969m3-1996m12), the paper finds that net exports rise by about 0.2% of GDP in the first six quarters after a one-percentage-point tightening, and that the average change in net exports to GDP over that horizon (0.19) closely matches the model’s own short-run prediction, with the initial response dominated by a fall in imports consistent with the model’s real-income and low-substitution-elasticity logic rather than by the expenditure-switching effect that dominates at longer horizons (Sec. 6, “Empirical evaluation of spillover channel,” Fig. 21).

The energy share of GDP is the key state variable: doubling or halving the paper’s baseline energy share (holding all other calibration fixed) roughly doubles or halves the size of both the direct output/inflation response to an energy shock and the magnitude of the monetary-policy spillover externality across countries, and the responses scale more than proportionally because a higher energy share also lowers the average elasticity of substitution χ (Sec. 7, Fig. 22-23). The paper’s overall policy conclusion is that any individual energy-importing country’s monetary tightening is a costly and limited tool against energy-driven inflation on its own, but generates a positive externality when coordinated across importers; fiscal policy, conversely, is a powerful and comparatively cheap tool for shielding an individual country, but generates a negative externality on other importers, especially when implemented as price subsidies – pointing the authors toward a policy mix of aggressive, coordinated monetary tightening paired with fiscal relief targeted at the poor, while explicitly avoiding energy price subsidies (Sec. 1, 8).