Short-Term Tax Cuts, Long-Term Stimulus NBER
- Date Posted:
- Is Database:
- Database
Corporate income tax cuts lead to significant & persistent increases in R&D spending & productivity, with effects peaking around 8 years post-shock. In contrast, personal income tax cuts yield large but short-lived boosts to output & productivity.
Key Finding, "..Do transitory changes in corporate and personal income taxes have persistent effects on output? And what are the channels?The effects of personal income taxes on output are large and significant during the first six quarters but return to zero within two years. In contrast, a corporate income tax cut generates a smaller stimulus on impact but has larger expansionary effects over the medium- and long-term, with peak effects occurring around eight years after the shock.Variable factor utilization is crucial to produce the short-run response of productivity. While the dynamic effects of personal income taxes are short-lived on productivity and negligible on R&D expenditure, changes in corporate income taxes are associated with a sustained, but temporary, rise in R&D spending and a persistent increase in aggregate productivity. R&D is key for the model’s ability to generate a persistent response of productivity and output following a temporary change in corporate income taxes. A standard New-Keynesian model would not deliver pro-cyclical productivity, thereby missing the estimated output responses to either shock...."
The Evidence
“…Using the approach outlined above, we now present our baseline set of empirical estimates about the longer-term effects of cuts to corporate and personal income taxes. We first focus on the response of the average tax rate, GDP and productivity, with the latter being a key and novel focus of our analysis. We will then extend our empirical evidence to examine investment, R&D and consumption to shed light on the most likely mechanism driving the GDP responses. Each further variable is added to the benchmark data vector Z one at the time to avoid a sharp increase in the number of parameters to be estimated. In Figure 1, we present our first set of main results. The figure contains two columns. On the left, we show the IRFs to a reduction in the average corporate tax rate. On the right, we show the results for a reduction in the average personal tax rate. The impact effect is normalized so that both shocks reduce their respective average tax rate by 1 percentage point in the first period. The solid red lines are the posterior medians and the shaded bands refer to 68% and 90% (Bayesian) credible intervals. Impulse response functions are computed using posterior draws of the coefficients A0 and B1. The solid blue lines come from the estimated structural model that will be presented, solved and estimated in Section 5. The first row in Figure 1 reveals that, following a shock to corporate and personal income taxes, the average tax rates decline temporarily. The change in the average corporate tax rate (first column) loses significance after 8 quarters and goes back to zero after around 20 quarters. The changes in the average personal income tax rate are somewhat less persistent, losing significance after 6 quarters and touching zero after around 16 quarters. Despite the different method (i.e. local projections vs. VAR), these results largely replicate the findings in Figures 2 and 3 of Mertens and Ravn (2013), where the results are plotted for the first 20 quarters. In short, the estimated tax cuts are rather transitory….”
“…The second row in Figure 1 shows the impulse response functions for the percentage response of real GDP. As expected, the IRFs for the first 20 quarters are very comparable to the main figures in Mertens and Ravn (2013). What is new is our estimate of the longer-term effects beyond quarter 20. Looking at the first column it is clear that, despite the transitory nature of the corporate tax reduction, there are very persistent effects on real GDP, whose short-run increase of 0.5% persists throughout the ten year period shown in the figure. In other words, the corporate income tax cut has disappeared after 5 years, but the effect on the level of economic activity is still sizable and significant after 8 years. The second column, however, reveals that the average personal tax rate cut does not produce such long-lasting dynamics.The underlying personal income tax cut is only slightly more transitory than the corporate tax cut but its effects on GDP are far less persistent and appear to die out already after two to three years after the shock hits. A similar picture emerges for productivity, as shown in the third row of Figure 1. Both tax rate cuts boost productivity on impact, with the size of the initial response to a personal income tax cut being much larger than for a cut to corporate taxes. On the other hand, the effects of corporate tax cuts grow over time and remain significant even after 10 years. In sharp contrast, the response of productivity to a change in personal income tax rates is not statistically different from zero already after two years. The clear difference in the short-run and long-run properties of the two taxes can be illustrated further by looking at the joint posterior distribution of transitory and persistent responses of the variable of interest in Figure 2.The top row refers to GDP while the bottom row represents productivity. We use ‘transitory’ (or shorter-term) to mean the effect estimated within 2 years after the shock, while ‘persistent’ (or longer-term) represents the dynamic effects estimated beyond the two year horizon The horizontal axis shows the response to a corporate tax cut, while the vertical axis shows the associated responses to a personal income tax cut. For the sake of exposition, each panel also reports the 45o degree line, which is the locus of points where the effects of personal and corporate income taxes on the variable of interest are numerically identical. Posterior draws below (above) the 45o degree line indicate a larger impact of corporate (personal) taxes. The share of draws below the 45o degree line, which we denote as δ in Figure 2, can therefore be seen as a measure of the probability that corporate tax changes have larger effects. In Appendix Figure A.2, we show that our flat priors for the LP parameters imply very disperse joint prior distributions for the short-run and long-run effects of the two shocks on GDP and productivity. These joint prior distributions are centered at (0,0). As a result, there is an even chance that the (persistent and transitory) effects of one type of tax will dominate the other. The left column indicates that for only about 6% of posterior draws do corporate income tax changes have larger short-term effects on GDP (top row) and productivity (bottom row) than personal income tax changes. In sharp contrast, the right column of Figure 2 reveals that corporate income tax cuts have significantly larger long-term effects than personal income tax cuts in about 93% of posterior draws for GDP and 98% for productivity. We conclude that the evidence of heterogeneous responses across both forecast horizons and type of income tax in Figure 1 is significant at conventional levels…”
“…In Appendix A3, we report the responses of Total Factor Productivity (TFP), total hours and employment. Our theoretical model will not, however, feature an extensive margin so the employ- ment IRF will not be used in the structural estimation in Section 5. The three main takeaways from this additional analysis are that: (i) the response of total hours and employment to a corporate income tax cut is typically modest and insignificant; (ii) in contrast, changes in personal income taxes have a significant but short-lived impact on total hours but muted effects on employment; (iii) the effects on TFP are very similar to those based on labour productivity. In summary, in the short-run, the effects of personal income tax changes on output, productivity and hours worked are significantly larger than the effects of corporate income tax changes. Over longer horizons, however, the responses of output and productivity to corporate income tax cuts are large and highly significant. In contrast, the long-run effects of a personal income tax cut are indistinguishable from zero, both in economic and statistical terms. In the next section, we will extend our empirical analysis to investment, R&D expenditure and consumption in an effort to shed light on the possible mechanism behind the heterogeneity documented in Figure 1 and Figure 2….”
The Mechanisms
“….The impulse responses in Figures 1 and 2 replicate the results in Mertens and Ravn (2013) over the first 20 quarters, which is the horizon at which most previous literature stops. On the other hand, we have shown that there are significant longer-term effects of corporate tax cuts that persist beyond the typical IRF horizons presented in earlier work. These persistent dynamics are not, however, evident for the effects of personal tax cuts. In the rest of the paper we investigate what may explain these findings. In this sub-section, we look at a number of additional variables that could offer insights on the transmission mechanism, especially at longer horizons. These are R&D expenditure, investment and consumption expenditure. The endogenous growth literature argues that R&D spending has the potential to generate persistent effects on both output and productivity. On the other hand, studies in the Real Business Cycle tradition emphasize the role of capital expenditure as an important propagation mechanism. Finally, given such longer-term output responses, we would also expect to see persistent effects on household expenditure for corporate tax changes, with short-lived effects from personal tax changes. The findings are reported in Figure 3. The first row shows the impulse responses of R&D expenditure to a corporate tax cut (left column) and to a personal tax cut (right column). The second and third rows show the dynamic effects on investment and consumption respectively. Red lines represent medians and 68% credible sets of the impulse response posterior distributions. Shaded areas refer to 90% central intervals. As discussed in Section 2, each variable is added one at the time to our baseline dataset to avoid a sharp increase in our already richly parameterized local projections….”
“…The evidence in the first row of Figure 3 suggests that the effects of corporate tax cuts (first column) on R&D are initially negligible but become significant at about one year after the shock. The increase is persistent and reaches a peak of 1.4% at quarter 18 before returning to zero after nine years. The effect also loses significance after six years. The response of investment to corporate tax changes (second row) is equally strong but its significance seems more short-lived. Finally, the consumption profile (third row) is similar to the pattern of the impulse responses of output and productivity in Figure 1.The significant and sustained rise in R&D seems a plausible candidate for explaining the persistent increase in productivity reported in Figure 1. In the next sections, we will explore this conjecture formally by developing and estimating a structural model with endogenous growth via R&D. The estimated effects of a personal income tax cut (second column of Figure 3), paint a different picture. The response of R&D is never statistically different from zero while the change in investment is larger but far more transitory than for corporate tax changes. The effects on R&D and capital expenditure suggests that the sharp and short-lived increase in productivity after a personal income tax cut in Figure 1 does not come from firms’ innovation activities. Later, we show that this is consistent with a short-run labor utilization story. Finally, the response of consumption in the bottom row largely inherits the shape of the GDP profile, as was the case for corporate taxes. This is consistent with the notion that corporate taxes raise labor income persistently, while personal taxes affect incomes only temporarily. In summary, the evidence in this section is consistent with a transmission mechanism in which R&D responds to a corporate tax shock (but not to a personal tax shock) and this triggers an endogenous response of productivity, which in turn drives a persistent effect on GDP. In Appendix Figure 5, we provide further support for this interpretation by looking at sectoral real gross output from the U.S. Bureau of Economic Analysis’s Industrial Accounts. We classify sectors in two groups based on their R&D intensity and estimate the heterogeneous effects of corporate and personal tax cuts. The estimates reveal that the output response to corporate tax changes is significantly larger in sectors with high R&D intensity. In contrast, there is no statistical difference in the output responses of the two groups of sectors to personal tax changes….”
James Cloyne, Joseba Martinez, Haroon Mumtaz and Paolo Surico, “Short-Term Tax Cuts, Long-Term Stimulus,”National Bureau Of Economic Researchhttps://www.nber.org/papers/w30246