1. Local Projections and VARs Estimate the Same Impulse Responses
Citation: Plagborg-Møller and Wolf [2021]. Econometrica 89(2):955-980, 2021.
Research question: Do local projections (LP) and structural vector autoregressions (SVAR) identify the same impulse responses under common assumptions?
Identification strategy: Formal theoretical proof under a general linear time-series model. The authors show that LP and correctly specified SVAR are numerically equivalent in population, and that in finite samples, LP is strictly more robust to misspecification of dynamics.
Key result: LP and SVAR estimate the same population impulse responses when the SVAR is correctly specified. When the SVAR lag order is misspecified (truncated), LP with enough lags is consistent while the misspecified SVAR is not. The trade-off is efficiency: correct SVAR is more precise than LP in finite samples, but the efficiency gain vanishes when the SVAR is even slightly misspecified.
Takeaway: LP is not just an alternative to SVAR—it is a robust semi-parametric version of it. Researchers should default to LP when uncertain about the correct dynamic model, and use SVAR only when structural identification requires the system approach.
2. Fiscal Stimulus in a Monetary Union: Evidence from US Regions
Citation: Nakamura and Steinsson [2014]. American Economic Review 104(3):753-792, 2014.
Research question: What is the causal effect of government spending on output? What is the size of the fiscal multiplier?
Identification strategy: The authors use variation in US military procurement spending across states as a quasi-experimental instrument. Military spending fluctuates with national defence decisions, creating differential spending shocks across states that are plausibly exogenous to local economic conditions. They compare state-level output growth in high-military-spending states to low-military-spending states during periods of spending increases.
Key result: The open-economy relative multiplier is approximately 1.5: a $1 per capita increase in government spending raises per capita income by $1.50 in the affected state. This exceeds the closed-economy multiplier because it does not account for regional leakages (imports from other states). Structural model calibration suggests a closed-economy multiplier of around 1.0-1.5.
Takeaway: The paper provides clean causal evidence that the fiscal multiplier is positive and meaningfully large, resolving a longstanding debate between RBC models (multiplier < 1) and Keynesian models (multiplier > 1) in favour of the Keynesian view, at least at the regional level.
3. The Macroeconomic Effects of Tax Changes: The Romer-Romer Narrative Approach
Citation: Romer and Romer [2010]. American Economic Review 100(3):763-801, 2010.
Research question: What are the dynamic macroeconomic effects of tax changes?
Identification strategy: Romer and Romer [2010] construct a narrative measure of exogenous tax changes: they read historical Congressional testimony, budget documents, and presidential messages to classify each major US tax change as either "endogenous" (responding to the business cycle) or "exogenous" (motivated by deficit reduction, long-run growth, or inherited from prior legislation). Only exogenous tax changes are used as shocks, plausibly avoiding the reverse-causality problem that plagues regressions of output on standard tax measures.
Key result: A tax increase of 1% of GDP reduces output by approximately 3% over three years, with the response peaking around 8-10 quarters. The multiplier is substantially larger than conventional estimates using unadjusted tax variation, consistent with the clean narrative identification removing endogenous variation that would attenuate the estimate.
Takeaway: The narrative approach to identifying policy shocks—now widely extended to monetary policy, oil price shocks, and immigration policy—is one of the most influential methodological innovations in empirical macroeconomics of the past two decades.
4. State-Dependent Effects of Fiscal Policy: Auerbach and Gorodnichenko (2012)
Citation: Auerbach and Gorodnichenko [2012]. American Economic Journal: Economic Policy 4(2):1-27, 2012.
Research question: Does the fiscal multiplier depend on whether the economy is in a recession or expansion?
Identification strategy: Nonlinear local projections with a smooth-transition function of the output gap. Structural government spending shocks are identified from forecast errors of professional forecasters (Blanchard-Perotti type identification). The recession/expansion state is defined via a seven-quarter moving average of GDP growth.
Key result: The cumulative fiscal multiplier at a 2-year horizon is approximately 2.5 in recessions and 0.5 in expansions—a five-fold difference. The recession multiplier is statistically significantly positive; the expansion multiplier is indistinguishable from zero in most specifications.
Takeaway: The state-dependent multiplier is now a benchmark result in fiscal policy analysis. It provides a causal rationale for counter-cyclical fiscal policy: stimulus is most effective precisely when the economy needs it most. The paper also demonstrated the practical value of nonlinear LP methods, spurring a large follow-up literature.
5. High-Frequency Identification of Monetary Non-Neutrality: The Information Effect
Citation: Nakamura and Steinsson [2018]. Quarterly Journal of Economics 133(3):1283-1330, 2018.
Research question: How large are the real effects of monetary policy shocks? Does the "information effect" bias high-frequency identification of monetary shocks?
Identification strategy: High-frequency identification: the authors measure monetary policy shocks as changes in federal funds futures in a narrow window around FOMC announcements (the Gürkaynak-Sack-Swanson methodology). They then distinguish two components of FOMC announcements: the pure monetary policy surprise (rate change conditional on the Fed's information set) and the "information effect" (the component that reveals information about the Fed's economic forecast, thereby shifting private-sector expectations).
Key result: After decomposing the shock, the pure monetary policy surprise has significantly larger real effects than conventional high-frequency identification suggests. The information effect component—positive rate surprises that reveal positive Fed forecasts—actually increases stock prices and output, the opposite of what a contractionary policy shock should do. Ignoring the information effect significantly underestimates the causal effect of monetary policy on real variables.
Takeaway: Central bank communication is itself a causal policy tool distinct from the interest rate setting. Researchers using high-frequency IV for monetary policy must account for the information channel, or their instruments conflate the policy shock with the information revelation.
The papers reviewed above represent a selection of seminal and influential contributions to causal identification in macroeconomics. Full citations appear in the reference list.
References
- Auerbach, A. J. and Gorodnichenko, Y. (2012). Measuring the output responses to fiscal policy. American Economic Journal: Economic Policy, 4(2):1-27.
- Nakamura, E. and Steinsson, J. (2014). Fiscal stimulus in a monetary union: Evidence from US regions. American Economic Review, 104(3):753-792.
- Nakamura, E. and Steinsson, J. (2018). High-frequency identification of monetary non-neutrality: The information effect. Quarterly Journal of Economics, 133(3):1283-1330.
- Plagborg-Møller, M. and Wolf, C. K. (2021). Local projections and VARs estimate the same impulse responses. Econometrica, 89(2):955-980.
- Romer, C. D. and Romer, D. H. (2010). The macroeconomic effects of tax changes: Estimates based on a new measure of fiscal shocks. American Economic Review, 100(3):763-801.
- Stock, J. H. and Watson, M. W. (2018). Identification and estimation of dynamic causal effects in macroeconomics using external instruments. Economic Journal, 128(610):917-948.