An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output

Questions & answers

Q1. What is the paper’s identification strategy, and how does it differ from identifying monetary policy shocks?

The authors identify fiscal shocks using “institutional information about the tax and transfer systems and the timing of tax collections to construct the automatic response of fiscal policy to economic activity, and, by implication, to identify changes in fiscal policy,” reasoning that because “decision and implementation lags in fiscal policy imply that… there is little or no discretionary response of fiscal policy to unexpected contemporaneous movements in activity” within a quarter, quarterly data lets them attribute the within-quarter correlation between taxes/spending and output almost entirely to automatic stabilizers rather than discretion. This is explicitly the opposite assumption from Bernanke-Mihov-style monetary SVARs, which assume private-sector variables do not react contemporaneously to policy: Blanchard and Perotti instead assume “economic activity does not affect policy, except for the automatic feedback built into the tax code and transfer system,” closer in spirit to Sims and Zha (1996).

Q2. How, concretely, are the automatic elasticities of taxes and spending constructed?

The authors “could not identify any automatic feedback from economic activity to government purchases of goods and services; hence, we take b1 = 0,” so all of the identifying work falls on the tax side: net taxes are decomposed into categories (indirect taxes, personal income tax, corporate income tax, social security taxes), and for each category the within-quarter output elasticity is built as the product of “the elasticity of taxes of type i to their tax base” and “the elasticity of the tax base to GDP,” drawing on OECD-style institutional estimates and the authors’ own collection of statutory collection-lag and quarter-dependence features of the U.S. tax code (detailed in the Appendix).

Q3. What are the paper’s baseline (non-event-specific) multiplier estimates for taxes and for spending, and how sensitive are they to the trend assumption?

For tax shocks, “output falls on impact by about 70 cents and reaches a trough 5 quarters out, with a multiplier… of 0.78” under the deterministic-trend (DT) specification, versus a larger trough of “-1.33… after 7 quarters” under the stochastic-trend (ST) specification – “the negative response of output is more pronounced under the assumption of a stochastic trend.” For spending shocks, “GDP increases on impact by 0.84 dollars, then declines and rises again, to reach a peak effect of 1.29 after almost 4 years” under DT, versus a smaller peak of 0.90 (reached immediately) under ST. Across these and other specifications, “depending on the specification, the spending multiplier is larger or smaller than the tax multiplier,” and the authors state plainly that “there is no consistent evidence” for the traditional-Keynesian prediction that it should always be larger.

Q4. What happens when the authors isolate a single, well-known historical tax event – the 1975 rebate – rather than relying on the full-sample VAR shock?

As a check on the VAR-based shocks, the authors separately trace the dynamic effect of a dummy variable for the 1975:2 tax cut, finding the output response “reaching a peak after 4 quarters with a multiplier of 0.75 under DT and 1.02 under ST,” a magnitude broadly similar to the VAR-estimated tax multiplier, which the authors read as corroborating evidence that the two approaches – structural VAR shocks and a single, well-documented event – are capturing the same underlying effect rather than artifacts of either method alone.

Q5. What is the “distinctly non-Keynesian” finding on investment, and why do the authors consider it surprising?

Breaking the output response into its GDP components, the authors find that “private consumption is consistently crowded out by taxation, and crowded in by government spending” – consistent with a simple Keynesian model – but that both tax increases and spending increases produce “a considerable crowding out of investment,” which “implies a strong negative effect on investment of a balanced-budget fiscal expansion.” Under DT, a spending shock raises consumption by “0.50 (50 cents) on impact,” building to a peak of 1.26, while investment “declines for the first 5 quarters, with a peak crowding out effect of -1.0”; a tax shock lowers consumption by a peak of “0.35 (35 cents) after 5 quarters” and lowers investment by “0.36 on impact.” The authors note this combination – spending and taxes moving investment in the same direction rather than opposite directions – is “consistent with a neoclassical model, but inconsistent with a standard Keynesian approach,” in which the output effect and the interest-rate effect of a spending increase should instead push investment in opposite directions depending on their relative strength, and they call for further investigation rather than offering a resolution.

Q6. How robust is the tax multiplier to excluding different decades, and what does this imply about stability over time?

The authors report that “the exclusion of the eighties causes a substantial drop in the magnitude of the tax multiplier, under both ST and DT,” with the instability “more pronounced under DT: the tax multiplier ranges from -0.49 when the eighties are excluded to -1.45 when the nineties are excluded.” They similarly flag “some instability in the spending multiplier, in particular under ST,” without resolving why particular decades carry disproportionate weight – a caveat about subsample stability that qualifies how literally the point-estimate multipliers should be read as constant structural parameters.

Q7. How does the paper use the Korean War episode, and what does the defense/non-defense breakdown add?

Extending the sample back to 1949 to include the Korean War buildup, the authors estimate that “the maximum increase in output is more than 2.5 times the maximum increase in spending” under DT (1.5 under ST) during this early-1950s episode, larger than in the post-1960 sample; splitting spending into defense and non-defense components over the extended sample, they find “large and similar multipliers for defense and non-defense spending – 2.50 and 2.67, respectively,” under DT. The authors use this exercise partly to critique narrative/event-study estimates of defense-spending multipliers by other authors (Ramey and Shapiro 1998; Edelberg, Eichenbaum and Fisher 1998), showing that the implied multiplier for the same underlying data can be reported as “about 28” under one definition (peak output response over the initial spending shock) or “about 1.4” under another (peak-to-peak ratio) – illustrating, in the authors’ words, that how a “multiplier” is defined can by itself generate an order-of-magnitude difference with no change in the underlying estimates.

Q8. What are the paper’s own stated conclusions and acknowledged loose ends?

The authors summarize their main findings as: “When government spending increases, output increases; when taxes increase, output falls, " and “in most cases the multipliers are small, often close to one.” They explicitly flag that “the responses of investment and imports are difficult to reconcile with most macroeconomic theories,” and “while we do not attempt an explanation here, we believe this certainly deserves further investigation” – the paper closes by naming its own central puzzle (investment’s response to both spending and taxes) as unresolved rather than papering over it.