July 27, 2006
HP-29
Remarks by Robert Carroll,
Deputy Assistant Secretary for Tax Analysis
U.S. Department of the Treasury
Before the National Economists Club
Thank you for the opportunity to discuss with you the Treasury Department's efforts on dynamic analysis.
Before I begin, let me first acknowledge several individuals who have contributed to the work on dynamic analysis at the Treasury Department. Two individuals within the Office of Tax Analysis have contributed significantly to this effort. Jay Mackie has made contributions over many years and Craig Johnson has been at the heart of this work over the past year. We have also benefited enormously from a collaborative effort with John Diamond and George Zodrow, both with Rice University and affiliated with the James Baker III Institute on Public Policy. Many will remember John from his time with the Joint Committee on Taxation. This work could not have proceeded without the significant contributions of all of these individuals.
Earlier this week the Treasury Department released a report that details its dynamic analysis of permanent extension of the President's tax relief. This analysis was summarized in a box included in the Administration's Mid-Session Review released earlier this month.
This report represents a continuation of our work on dynamic analysis. As you may know, the Treasury Department also released a report on May 25th that summarized the dynamic analysis of the tax reform options prepared on behalf of the President's Advisory Panel on Federal Tax Reform last fall.
In many presentations on dynamic modeling, the presenter provides a detailed description of the model, focusing on the specific characteristics or embellishments that differentiate the model from previous work. For this audience, I think a more productive approach is to provide you with a brief overview of our approach, then focus on the key results - the main lessons, if you will - from our analysis, and then discuss what our next steps are at Treasury.
Modeling Approach
Back in February when describing our initiative to create a new dynamic analysis division at Treasury, we indicated that dynamic analysis would have the benefit of focusing attention on the broad economic effects of changes in tax policy. We also indicated that the effort would focus on the long-run effects of tax policy and acknowledged the results were dependent on financing and underlying assumptions. In this report, we have attempted to focus on these aspects of dynamic analysis.
It is also important to point out at the outset what our analysis does not do: It does not provide a dynamic score or estimates of the revenue feedback associated with the President's tax relief. This would involve translating the estimated changes in output into the associated change in revenues. We envision that dynamic analysis at the Treasury Department may ultimately evolve in that direction, just as it has to varying degrees at the Joint Committee on Taxation and the Congressional Budget Office, but we are still very much at the beginning of this effort.
The model we used is a conventional neoclassical growth model with overlapping generations of taxpayers - an OLG model. In this life-cycle model, tax policy affects the incentives to work, to save and invest, and to allocate capital among competing uses. It captures the intersectoral reallocation of capital that results from reducing the double tax on corporate profits, and the crowding out of private investment from financing the tax relief through issuing government debt. Representative consumers and firms incorporate future prices into their current period decisions of how much to save, work, and produce. Output is generated by four production sectors, and individual level decisions of representative consumers determine the aggregate level of labor supply and savings in each year. We included a simple representation of international capital flows. We also considered different assumptions for financing the tax cuts and considered how the results change with different values for underlying parameters.
While we used three different models to analyze the broad tax reform proposals put forward by the tax panel last fall, in the dynamic analysis of the President's tax relief we chose to use just one model, the overlapping generations model. This choice was made in large part because the version of the model we are working with is more detailed than the versions of the other models we used in the analysis conducted on behalf of the tax panel, and thus better suited for analyzing the specific features of the President's tax relief.
Five Lessons
The report provides five basic lessons:
1. Many claim that tax relief can increase economic growth, and this report supports this claim.
According to this analysis, the President's tax relief would increase real GNP by 0.7 percent in the long-run. In a $13 trillion economy, this amounts to an additional $90 billion, in today's dollars, each year, forever.
2. All tax changes are not created equal.
Some tax changes, such as the lower tax rates on capital gains and dividends, reduce the tax rates on capital, increasing the incentive to invest, and increasing incomes and living standards in the long run by making labor more productive.
Other tax changes, such as the lower tax rates on ordinary income, reduce tax rates on labor, which increase the after-tax reward to work, labor supply, and real GNP.
Yet other tax changes, such as the expanded child tax credit, marriage penalty relief and new 10 percent rate bracket, can provide other types of benefits. They may not encourage long-run growth by lowering tax rates on capital or labor, but they can provide important and timely stimulus to the economy in the near-term. This can be particularly important during a period of economic weakness, such as the U.S. economy faced several years ago. And, this is especially important when the monetary policy has already been used aggressively, as it was in 2001 and 2002.
Of course, tax changes can also be used to maintain or increase the progressivity of the income tax, and help families with their own economic challenges.
The Treasury analysis decomposed and separately considered the effects of three different portions of the President's proposal to permanently extend the tax relief:
1. Lower tax rates on dividends and capital gains;
2. Reduction in the top four ordinary tax rates; and
3. Expansion of the child tax credit, the marriage penalty relief, and the new 10 percent rate bracket.
The Treasury analysis finds that more than half of the 0.7 percent increase in long-run GNP is associated with the lower tax rates on dividends and capital gains, even though this policy change accounted for less than 20 percent of the static revenue loss. The increase in long-run GNP rises to 1.1 percent when the lower tax rates are added.
The rise in GNP is smaller when the remaining tax changes - the child tax credit, marriage penalty and new 10 percent rate bracket - are included. This result, at first glance, may seem counter-intuitive, but in this model these policies have - what economists call - income effects. Because these policies - the child tax credit, marriage penalty relief, and the new 10 percent tax rate - increase after-tax incomes, and have little effect on incentives, some taxpayers may respond by increasing their leisure and working less, which is the primary reason GNP is not as high in the long-run - 0.7 percent rather than 1.1 percent.
But again, in looking back at the 2001 and 2003 tax relief, it is crucial to remember that the broad policy objectives were two-fold: 1) to shore up and strengthen an economy that faced significant risks - a double dip recession, disinflation; and, 2) to promote long-run growth.
The package of policies accomplished those twin goals by accelerating the rate at which the economy returned to full capacity and, as this report suggests, promoting long-term growth.
3. In doing this work, it is very much our goal to be as transparent as possible in the underlying assumption and results. One key assumption in analyzing the long-run effects of the tax relief is how it is ultimately financed.
A key feature of the model used for this analysis is the recognition that the government faces what economists call an intertermporal budget constraint. This means that the present value of taxes is tied to the present value of government spending. In simpler terms, when taxes are reduced, other offsetting changes are needed.
In this report, we considered two financing options: 1) lower future government spending; and 2) higher future taxes. Under the first financing assumption - lower future government spending - we report the base results - an increase in long-run GNP of 0.7 percent. But under the alternative financing assumption - higher future taxes, long-run GNP would actually be lower by 0.9 percent.
What is the intuition behind this result? It is really quite simple. In a model where consumers are forward looking, higher future taxes discourage economic growth. But these results are suggestive of another basic point: Permanent extension of the tax relief, financed by spending restraint, will encourage economic growth. Alternatively, if the tax relief is, in effect, temporary and, in the aggregate, offset with higher future taxes, real long-term GNP can be expected to fall.
4. Also, in the interest of being as transparent as possible we also considered the sensitivity of the results to underlying assumption.
The parameters of the Treasury model are taken from the consensus of the professional literature, but they are not pinned down with certainty. The GNP estimate of 0.7 percent is the result from our base case simulation, but the report also shows that this estimate can range from 0.1 to 1.2 percent by changing the model's parameters within plausible ranges.
5. Finally, in contrasting the different policies, it is worth noting that the lower tax rates on dividends and capital gains increase GNP in the long-run, even when financed by higher future taxes. This policy change lowers some of the highest marginal tax rates that taxpayers face through the double tax on corporate profits. The future tax increases under this financing assumption are broad-based and, essentially, marginal tax rates decline on average, which lowers the efficiency cost of raising tax revenue. Or, put slightly differently, the higher future tax revenue needed to finance the temporary tax relief actually results in an overall lower tax burden as compared to current law because the high taxes on corporate profits are, in effect, replaced with broad-based taxes.
Future Improvements
While we view this report as a significant step forward, future work will continue to improve and refine the modeling. There are many aspects of the economy that are not captured in this analysis. Like all economic models, this model employs important simplifying assumptions, and other economic models, which make different assumptions, could yield different results. The model used in this analysis departs from economic reality in a number of important ways.
First, this model does not account for short-term deviations in output from potential GNP. This implies that the model does not capture some of the short-run benefits of tax relief when the economy is below its potential, such as with the 2001 and 2003 tax relief. As discussed in the report, a different model was previously used by Treasury to analyze the demand-side or stimulative effects of the tax relief in the near term.
Second, the treatment of international capital flows is quite simple. A broader model would employ a more sophisticated representation of these flows and would also include international trade in goods. To the extent the economy is more open to international capital flows, the effects of crowding out associated with higher government debt could be dampened.
Third, this model assumes perfect certainty and perfect competition. Of course, we live in a world with uncertainty. Also, some models with imperfect competition find that capital income taxes have larger distortion effects because they are, in effect, layered on top of the preexisting distortion associated with imperfect competition.
Finally, the financing assumptions used in the simulations are conventional for this type of analysis and are not meant to be predictions of what policies might actually occur. Numerous other policy prescriptions could also occur. For example, if the tax relief is financed through reductions in government spending that occur sooner than assumed in the simulations in the Treasury analysis, government borrowing would be less, the crowding of private investment would be smaller, and the increase in long-run output would be larger.
Nevertheless, we think it is a very good start.
Next Steps
What are the next steps? As I mentioned above, the Treasury Department is working to expand its capability for dynamic analysis by standing up a new Division of Dynamic Analysis within the Office of Tax Analysis. The additional resources associated with creating this new division will help us enhance and expand our existing capabilities to address some of the limitations listed above. The additional resources are also important to conduct this analysis on a more systematic and regular basis for broad policy changes. This is of immediate relevance to the Department's current effort to evaluate different approaches for reforming the tax system.
Since this initiative was announced in early February, I have grown to better appreciate the different perspectives brought to this issue. There are some who have a long-standing interest in this subject, but who are very concerned that we at Treasury may not do this the right way. Then there are others who are concerned that dynamic analysis may politicize the work we do at Treasury. To be clear, we at Treasury are well aware of the sensitivities that arise in discussions of dynamic analysis and we take these concerns very seriously.
Again, being as transparent as possible is very important to this endeavor. It allows the professional and policy community to understand and evaluate key assumptions and results. So we need to continue to clearly divulge those assumptions and release enough information regarding the models and the results so that those outside of Treasury can understand and evaluate the work we are doing. Reports, such as the one released earlier this week, help provide this transparency.
Sensitivity analysis is also critically important and will continue to be a central part of our work. As is clear from this report and the work in this area by CBO and the JCT, this type of analysis is sensitive to key assumptions, especially how a tax change is financed, so we need to continue to carefully consider those assumptions and report how the results vary with different sets of assumptions.
Continuing an open and continuing public dialogue on our work is also important. It would be clearly too much to expect that all will agree with every choice or assumption we have made. [The number of phone calls and emails I have gotten in the last day and a half can attest to this.] But the opportunity to discuss the work in forums like this and elsewhere is very helpful.
This work should be viewed very much as building on the continuing evolution of the manner in which we have integrated the behavioral aspects of taxation on economic decision making in all our work at Treasury. And, it should be remembered that we already exercise considerable judgment, and I would say good judgment, in the work we do on conventional revenue estimates and a variety of other analyses.
Finally, I think it is important to reflect that this type of analysis clearly places attention on the economic effects of tax policy, both in the long-run and over the transition path to this long-run. It helps frame the discourse on tax policy around these economic benefits, rather than the five or ten year budgetary effects of proposals. It also helps inform the discussion of tax policy by focusing attention on the key decisions that drive the results produced by this model. To what extent does a policy affect the incentives to invest or supply more labor, or work primarily through changes in the after-tax incomes of consumers? Over the longer-term, the success of dynamic analysis will largely be determined by how well it is communicated to the policy community, how open the process remains, and to what extent this type of analysis complements more conventional analysis of tax policy.
Again, I thank you very much for the opportunity to share some of these thoughts with you. I am happy to take your questions.