Fiscal Policy Shocks

Definition

A fiscal policy shock is the unexpected, exogenous component of government spending or taxation — the part not explained by the systematic response of fiscal policy to the state of the economy. Estimating the dynamic effects of such shocks on output, consumption, and investment (the fiscal multiplier question) requires separating exogenous policy changes from the endogenous response of taxes and spending to activity, which is the central identification problem of empirical fiscal-policy analysis.

Key Ideas

• The identification problem. Observed movements in taxes and spending mix (i) the automatic response of tax revenue and transfers to the business cycle (automatic stabilizers), (ii) the systematic discretionary response of policymakers to the economy, and (iii) genuinely exogenous fiscal shocks. Multipliers are only interpretable once (iii) is isolated.
• Institutional-elasticity identification (Blanchard-Perotti 2002). Use tax-code and transfer-system elasticities to compute the automatic response of net taxes and spending to output (the elasticity of net taxes to output ≈ 2; government purchases ≈ 0). Because at quarterly frequency discretionary policy cannot observe output and respond within the quarter, the systematic within-quarter response is assumed zero, so the remaining reduced-form residual identifies the structural fiscal shock — no recursive/Cholesky ordering required.
• Narrative identification (Ramey-Shapiro). An alternative that dates large exogenous shifts in government (especially military) spending from the historical record, treating those dates as the exogenous shocks — robust to within-sample endogeneity but coarse.
• Sign-restriction identification (Mountford-Uhlig 2009). Extend Uhlig's (2005) agnostic sign restrictions to multiple shocks: jointly identify a government-revenue and a government-spending shock by sign-restricting the fiscal variables and requiring both to be orthogonal to a sign-identified business-cycle and monetary-policy shock (which filters out automatic responses). Fiscal shocks then span a two-dimensional space, so deficit-spending, deficit-financed tax cuts, and balanced-budget expansions are different linear combinations — with responses of GDP/consumption/investment left unrestricted. Deficit-financed tax cuts deliver the largest GDP multiplier (≈5 at a five-year horizon).
• The fiscal multiplier. The cumulative output response per dollar of the fiscal shock. Blanchard-Perotti find a government-spending multiplier around one; magnitudes remain contested and state-dependent (recession vs. expansion, monetary accommodation, slack).
• Identification = a prior on elasticities (Caldara-Kamps 2017). The competing schemes above are near-observationally-equivalent: each is a different assumed value of the fiscal output elasticities, and the multiplier is an explicit analytical function of those elasticities. So the notorious cross-study dispersion in multiplier estimates is a disagreement about elasticities, not about the data — and disciplining the elasticities with extra-model (institutional/micro) information sharpens the multiplier posterior. For the U.S. 1947–2006 the spending multiplier is, if anything, larger than the tax multiplier at all horizons. This is the fiscal instance of treating identification as a prior on an interpretable structural parameter.
• Fiscal foresight / anticipation. Agents (and legislation) often see tax and spending changes coming before they are implemented, so the econometrician's information set lags the agents' — the VAR shock is partly anticipated, an invertibility/informational failure (Ramey 2011) that biases estimated responses.

How It Works

Blanchard and Perotti estimate a trivariate quarterly VAR in $(T_t, G_t, Y_t)$ — net taxes, government spending, output. The reduced-form residuals $(e^T, e^G, e^Y)$ relate to structural shocks $(\varepsilon^T, \varepsilon^G, \varepsilon^Y)$ through $$e^T_t = a_1 e^Y_t + a_2 \varepsilon^G_t + \varepsilon^T_t,\qquad e^G_t = b_1 e^Y_t + b_2 \varepsilon^T_t + \varepsilon^G_t,$$ where $a_1$ (the contemporaneous elasticity of net taxes to output) is fixed from institutional data rather than estimated, and $b_1 \approx 0$ (spending does not respond automatically within the quarter). With the endogenous-to-output channel calibrated, the fiscal shocks are recovered and their impulse responses traced. An ordering assumption between $\varepsilon^T$ and $\varepsilon^G$ (which decides first) is needed but shown to be immaterial.

Why It Matters

• The "Blanchard-Perotti approach" is the default SVAR identification in the vast fiscal-multiplier literature, and the empirical benchmark against which theoretical (RBC vs. New Keynesian) models are judged.
• Its consumption result — government spending raises private consumption — is a central empirical fact distinguishing Keynesian from neoclassical fiscal transmission.
• It is the fiscal analogue of monetary-policy-shock SVARs, showing how calibrated (rather than assumed-zero) contemporaneous restrictions can identify a policy shock.

Open Questions

• Multiplier size and state dependence. How large is the multiplier, and how much does it depend on the business cycle, the monetary regime, and the exchange-rate regime?
• Anticipation. How badly does fiscal foresight bias VAR-based estimates, and do narrative or proxy-SVAR (external-instrument) methods recover the truth?
• Composition. Do the effects differ across spending types (investment vs. consumption, defense vs. transfers) and tax instruments?

Related

• Structural Identification
• Sign Restriction Identification
• Bayesian SVAR Identification
• Monetary Policy Shocks
• Impulse Response Function
• Caldara-Kamps (2017) The Analytics of SVARs
• Dario Caldara
• Christophe Kamps
• Invertibility Problem
• External Instruments
• Blanchard-Perotti (2002)
• Mountford-Uhlig (2009)
• Olivier Blanchard
• Roberto Perotti
• Andrew Mountford
• Harald Uhlig