In recent years, governments have relied heavily on fiscal stimulus to navigate economic crises, such as the Covid-19 pandemic. This column uses historical US aggregate data and state-dependent local projections to analyse whether government spending multipliers are different for positive versus negative shocks. It shows that the GDP response to a spending increase is roughly proportional to the response to a spending decrease, contradicting earlier findings of asymmetric effects. The finding of no significant asymmetry in government spending multipliers has important implications for the costs of fiscal consolidation.
In recent years, governments worldwide have relied heavily on fiscal stimulus to navigate economic crises. From the fiscal policy responses following the 2008 financial crisis to the massive government spending during the COVID-19 pandemic, policymakers have debated the effectiveness of public spending in stimulating economic growth, or the size of the ‘government spending multiplier’ – how much GDP increases for every dollar the government spends.
A crucial question in this debate is whether the effects of government spending depend on the situation, for example whether the economy is in recession or whether spending is increasing or decreasing. Research in the last decade has investigated whether multipliers depend on the state of the economy (e.g. Auerbach and Gorodnichenko 2010, Ramey and Zubairy 2015, Castelnuovo and Caggiano 2015) or display other types of time-variation (e.g. Rossi et al. 2022). An older literature explored asymmetry, meaning whether declines in government spending have larger effects on economic activity than rises in government spending. For example, Hooker and Knetter (1997) found that military cutbacks had larger effects than military buildups on state economies, and Davis et al. (1997) found asymmetries in the effects of both oil shocks and government spending shocks on regional economies. More recently, work by Barnichon et al. (2022) considered aggregate data and argued that fiscal multipliers depend on the sign of the shock—negative shocks (spending cuts) lead to larger contractions than the expansions generated by positive shocks (spending increases).
If spending cuts reduce GDP more than equivalent spending increases boost it, then efforts to reduce budget deficits could have more severe economic consequences than previously thought. There would then be two costs of a rise in government spending: the standard cost of raising taxes to finance the spending, and the amplified negative effects of the wind-down in government spending. This asymmetry thus has potentially significant implications for designing fiscal policy, particularly in periods of consolidation.
In Ben Zeev et al. (2025), which is an expanded working paper version of Ben Zeev et al. (2023), we revisit this question of multipliers being different based on the sign of the shock. To motivate our scepticism of asymmetric effects on multipliers, we first review events during two influential episodes in the US historical data — the two world wars. Figure 1 shows the behaviour of real government purchases and real GDP, both divided by potential GDP. In each case, government spending rose when the US became involved in the war. At the end of the war, real government spending returned to its pre-war fraction of potential GDP in less than a year. GDP rose and fell along with government spending. If anything, GDP rose more with the rise in government spending than it fell with the decline in government spending at the end of the war. In both cases, the recession at the end of the war was shallow and brief. Thus, there is no evidence in the raw data suggesting that declines in government spending have greater effects than rises in government spending.
Figure 1 Case study of two world wars
Note: Real government purchases (G, dashed) and GDP (Y, solid) divided by potential GDP.
Source: Ramey and Zubairy (2018) data.
Our paper exploits historical US aggregate data spanning more than a century and employs state-dependent local projections to address whether government spending multipliers differ by the sign of the shock. The use of long historical series ensures that our sample includes multiple episodes of large government spending variations, such as the military expansions in the two world wars, the Korean War and other major buildups, and subsequent drawdowns. To identify a government spending shock that is not only exogenous to the state of the economy but is also unanticipated, we use Ramey and Zubairy’s (2018) military news series from 1889 to 2008. This news series focuses on changes in government spending that are linked to political and military events, since these changes are most likely to be independent of the state of the economy.
In the statistical analysis, we first use diagnostic tests from Ben Zeev (2020) to check for nonlinearities in the data and find that both the sign and size of shocks to government spending matter for impulse responses of spending and GDP to the shock. We then apply state-dependent local projections to examine how GDP reacts to increases versus decreases in government spending, where we model asymmetry as a state in the framework of Ramey and Zubairy (2018). While we find that the impulse responses of GDP and government spending differ depending on whether the shock is positive or negative, these differences do not translate into asymmetric multipliers. Figure 2 shows the cumulative multipliers to a positive shock (blue) and a negative shock (red). In other words, the GDP response to a spending increase is roughly proportional to the response to a spending decrease, contradicting earlier findings of asymmetric effects.
Figure 2 Cumulative multipliers by sign of shock
Note: 95% confidence bands based on HAC standard errors.
- Overall, we find no evidence of an asymmetric multiplier.
- We do find evidence of asymmetry in how government spending and GDP respond to fiscal shocks. Both government spending and GDP respond more strongly to a negative military news shock than a positive one, but — crucially — they do not translate into spending multipliers that differ by the sign of the shock. Because both the numerator (the GDP response) and the denominator (the government spending response) scale up by similar amounts, the multiplier itself does not change.
One reason we arrive at different conclusions than Barnichon et al. (2022) is the use of a state-dependent local projection approach, which provides more precise estimates of fiscal multipliers. The earlier study relied on functional approximations to impulse responses of government spending and output, which may introduce biases, particularly when combined nonlinearly to create cumulative multipliers. Additionally, we find that grouping zero changes in spending with positive shocks (rather than negative ones) can significantly influence the estimates, highlighting the sensitivity of results to methodological choices.
Concluding thoughts
The debate over fiscal multipliers remains central to economic policymaking, especially in an era of high debt and fiscal uncertainty. While earlier research suggested that government spending multipliers are asymmetric, new evidence calls this into question. In Ben Zeev et al. (2023, 2025), we provide evidence that fiscal multipliers do not differ by the sign of the shock.
As fiscal debates continue, it is crucial for policymakers to ground their decisions in rigorous empirical evidence. The finding of no significant asymmetry in government spending multipliers has important policy implications for the costs of fiscal consolidation.
Source : VOXeu
