The global recession of 2008-09, one of the longest and deepest since the Great Depression, has made the efficacy of fiscal-stimulus packages one of the most prominent policy debates in economics today. These packages typically attempt to smooth out business-cycle fluctuations through a combination of increased government purchases of goods and services (to replace falling private demand) and tax cuts or rebates. The spending component of these packages is typically motivated by the belief that the expenditure multiplier is greater than one – in other words, that total output in the economy will expand by more than the increase in government purchases. Unfortunately, the existing empirical evidence on the size of the spending multiplier is limited at best. In ongoing research (Barro and Redlick, 2009), we attempt to estimate the impact of changes in spending and taxation on economic output using long-term macroeconomic data from the US. We hope that our estimates from this project can provide a useful benchmark in thinking about the design and effectiveness of fiscal-stimulus programs.
Previous research has taken a variety of different approaches in attempting to measure the impact of fiscal policy tools on output. One approach, exemplified by Blanchard and Perotti (2002), has been to use vector-autoregression (VAR) models in which identifying assumptions are made on the order in which the variables are allowed to move. Typically, the government expenditure variable is allowed to move first, and the responses of other variables are treated as causal. Another approach comes from Romer and Romer (2008). Their project takes a “narrative” approach, in which they read the legislative record for evidence on the motivation behind tax changes, as well as the size of the intended impact on federal tax revenue. They use this evidence to categorise tax changes as either endogenous or exogenous and then measure the impact on output from the exogenous shifts.
Our own empirical work begins by extending the times series of average marginal income-tax rates in the US constructed by Barro and Sahasakul (1983, 1986). Figure 1 plots the results; details of the computations can be found in Barro and Redlick (2009). Conceptually, our measure of the average marginal tax rate corresponds to the substitution effect of taxes on incentives to work, produce, and invest. This construct differs from the Romer and Romer measure of prospective federal tax revenue (as a ratio to GDP), which captures the income effect of taxes on disposable income.
Figure 1. Average marginal tax rates in the US, 1912-2006
Our empirical analysis treats the dependent variable as the annual growth rate of real per capita GDP, with the change in government defence purchases and the lagged changed in the average marginal tax rate (along with other controls) as the explanatory variables. In samples that include World War II, we find the multiplier effect of defence purchases is precisely estimated (and highly statistically significant) in a range of 0.6 to 0.7. Advocates of fiscal stimulus often claim the multiplier may be higher when there is more slack in the economy, so we also include the interaction between the change in defence spending and the lagged deviation of unemployment from its long-run median. As predicted, the estimated coefficient is positive and significant, with the defence purchases multiplier rising by about 0.1 for every two percentage points by which the unemployment rate exceeds its long-run median of 5.6%. Even by this estimate, however, the defence-spending multiplier would reach one only at an unemployment rate of about 12%. While we note that it would be highly desirable to also estimate the effect of non-defence purchases on output, the absence of good instruments leaves us with no solution to the obvious endogeneity problem. This is an especially thorny issue since state and local outlays have come to dominate non-defence expenditure, and these outlays are particularly sensitive to fluctuations in tax revenue (which are in turn partly determined by aggregate economic conditions).
A further question of potential policy interest is the mechanism by which the expenditure multiplier falls below one. As an accounting identity, in order for total economic output to rise less than one-for-one with increased government spending, some other component(s) of GDP must decrease. A relevant question, therefore, is exactly what gets “crowded out.” We attempt to answer this question by breaking down the impact of increased government expenditure on GDP into its various components. In particular, we consider the impact on private consumer expenditure, gross private domestic investment, and net exports. Empirically, we find that the effect on private consumer spending is not significantly different from zero, while net exports are slightly (but significantly) crowded out and private investment falls considerably.
Increased government spending is often accompanied by tax cuts in most fiscal stimulus packages. Our analysis of the effects of tax changes on output involves entering our newly constructed measure of the average marginal tax rate by itself and in combination with the Romer and Romer tax variable. We believe our research complements theirs and hope that the combination of the two measures may allow us to distinguish between the substitution effect of tax-rate changes and the income effect of total tax liability changes. In a sample beginning in 1950 (using only our marginal tax-rate series), we find that a one percentage point decrease in the first lag of the average marginal tax rate produces a 0.6% increase in the annual growth rate of real per capita GDP. Unfortunately, this effect is harder to pin down in longer samples, particularly when they include World War II and the Great Depression. The fact that the Romer-Romer series and the changes in our measure of the average marginal tax rate from the federal income tax are correlated to the tune of 0.76 from 1950 to 2006 makes separating their effects difficult. It might be that income effects are nil (given the tax-rate effects), but it is also possible that the tax-rate effects are nil (given the effects on government revenue).
Our bottom line from this research is that a healthy scepticism is warranted when policymakers claim government-spending multipliers in excess of one. Our estimates suggest that the multiplier effect of defence spending falls more in the range of 0.6 to 0.8, and we find it unlikely that non-defence multipliers would be larger. Therefore, our conclusion is that total economic output increases less than one-for-one with increased government purchases. However, we do find evidence to support the view that tax cuts stimulate total output, with a one percentage point decrease in the average marginal tax rate leading to an increase of about 0.6% in the growth rate of real per capita GDP. As such, our preference in the design of fiscal stimulus packages would be for more tax cuts and less reliance on increased government spending.
Barro, R.J. and C.J. Redlick (2009). “Macroeconomic Effects from Government Purchases and Taxes,” unpublished, Harvard University, October.
Barro, R.J. and C. Sahasakul (1983). “Measuring the Average Marginal Tax Rate from the Individual Income Tax,” Journal of Business, 56, October, 419-452.
Barro, R.J. and C. Sahasakul (1986). “Average Marginal Tax Rates from Social Security and the Individual Income Tax,” Journal of Business, 59, October, 555-566.
Blanchard, O. and R. Perotti (2002). “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output,” Quarterly Journal of Economics, 117, November, 1329-1368.
Romer, C.D. and D.H. Romer (2008). “A Narrative Analysis of Postwar Tax Changes,” unpublished, University of California Berkeley, November.