CRS REPORT LOOKS AT MACROECONOMIC EFFECTS OF CAPITAL GAINS CUT.

                          Brian W. Cashell

This study presents the results of three econometric simulations of the potential effects of a capital gains tax cut using the Data Resources, Inc. (DRI) economic model. This model tends to have a fairly pronounced response to the change in the cost of capital. The simulations are based on Joint Tax Committee estimates of the revenue effects of the thirty percent exclusion.

The first simulation assumes that there is no change in the cost of capital, the second assumes a 0.5 percent drop in the cost of capital, and the third simulation assumes a 1.5 percent drop in the cost of capital. The 0.5 percent drop is consistent with our direct estimates of the effect of the capital gains tax, and would be the upper limit of any economic response as it assumes that the capital gains tax cut is instantly translated into an increase in investment. The 1.5 percent drop is consistent with measures used in another study, which we believe to be overstated because of failure to take into account the forgiveness of capital gains taxes at death in calculating the initial impact of the tax.

These simulations indicate that the effect of a capital gains tax cut depends on the translation of the capital gains tax change into an investment stimulus and the subsequent effects on the cost of capital. Over the first five years the first simulation results in a cumulative increase in real GNP of $7.2 billion. In the second simulation cumulative GNP is $22.4 billion higher, and in the third simulation cumulative GNP is $47.4 billion higher. As noted in the text, it is also possible that the capital gains tax cut could be contractionary, if it increased savings but not investment.

Although the estimated effect of the capital gains tax cut varies with the modeling of the cost of capital, the effects are very small and suggest that the capital gains tax cut will have no noticeable impact on the overall performance of the economy.

CONTENTS

BACKGROUND

SIMULATION RESULTS

CONCLUSIONS

POTENTIAL ECONOMIC EFFECTS OF A CAPITAL GAINS TAX CUT

Congress is considering a number of proposals to promote economic growth. Prominent among them is a thirty percent capital gains tax exclusion. This report presents the results of three econometric simulations estimating the potential effects of a thirty percent capital gains tax exclusion on output and employment (using Joint Committee on Taxation revenue estimates).

BACKGROUND

These simulations were done using the Data Resources, Inc. (DRI) compact macroeconometric model. This model is not structured in such a way as to allow changing tax rates applicable to capital gains. In order to simulate the potential economic effects of reducing capital gains taxes several adjustments were imposed on the model. In all three simulations personal tax receipts were changed to reflect Joint Tax Committee estimates of the revenue effects of the tax cut. These revenue effects are shown in table 1.

                               TABLE 1.

Using the November 1991 DRI TRENDLONG forecast of the U.S. economy as a baseline for these simulations personal tax receipts were adjusted by the amounts shown in table 1. In order to carry these simulations out to 1999 it was assumed that beyond fiscal 1994 revenue losses would be $3.0 billion annually.

SIMULATION RESULTS

What distinguishes each of the three simulations from the other is the effect of the tax cut on the cost of capital. Table 2 presents the results of the simulation which assumes no change in the cost of capital. This approach treats the capital gains tax cut the same as any other cut in personal income taxes. In this case the tax cut results in a maximum reduction in the unemployment rate of 0.03 percentage points, relative to the baseline. That is, the number of jobs would increase by approximate 3/100 of one percent. This peak effect is reached in 1994. Over the first five years real GNP is higher by a cumulative $7.2 billion.

If the capital gains tax cut were to increase the supply of saving the cost of capital would fall and stimulate additional investment spending. The Congressional Budget Office (CBO) in their study of the capital gains tax cut, using the Washington University Macroeconomic Model (WUMM), captured this effect by entering the capital gains tax change in the user cost of capital. This econometric model produced a 1.5 percent drop in the cost of capital. ¹

In a Congressional Research Service (CRS) analysis, Jane Gravelle suggested that the cost of capital would likely fall by about 0.5 percent. ² (This fall in the cost of capital is the maximum which could occur, and effectively assumes an infinitely elastic supply of savings.) Conversations with authors of the Congressional Budget Office study suggest that the econometric model used by CBO appears to have overstated the effect of the capital gains tax cut by not accounting for a large fraction of capital gains that are not taxed because they are passed on at death. CBO's own calculations of the potential effect of the capital gains tax cut on the cost of capital were similar to the CRS estimates. ³

Table 3 presents the results of the simulation assuming that the cost of capital falls by 0.5 percent in response to a capital gains tax cut (this change was directly entered into the cost of capital formula, since the DRI model does not include a capital gains tax lever). In this case the tax cut results in a maximum reduction in the unemployment rate of 0.05 percentage points, again in 1994. By 1999, however, the unemployment rate rises above baseline levels. Presumably, this is because a lower cost of capital would encourage firms to employ more capital and less labor than they would have otherwise. In this simulation, real GNP was higher than baseline levels by a cumulative $22.4 billion over the first five years.

Table 4 presents the results of the simulation in which the cost of capital is assumed to drop by 1.5 percent. As noted above, this simulation is based on an overstated measure of the effect of the cost of capital and is prepared for comparison purposes with the CBO simulation. This assumption yields a maximum reduction in the unemployment rate of 0.1 percentage points relative to the base case in 1994. As was the case in the previous simulation, unemployment is higher than in the baseline by 1999.

In all three simulations the peak stimulative effects of the tax cut were not realized for nearly two years.

In none of the simulations were feedback effects from taxes on increased output in the economy large enough to offset the revenue cost. The static revenue cost was $12 billion over the first five years and $21 billion over the entire eight year period. ⁴ The basic simulation reported in Table 2 would produce a $10.9 billion loss over the first five years and a $22.2 billion loss over the entire eight year period. Incorporating the cost of capital effect as in Table 3 produces a loss of $7.6 billion in the first five years and $14.8 billion over the entire period. ⁵

It may be of some interest to compare these simulations with those in other models. The table 4 simulations using the same cost of capital as CBO (a 1.5 percent reduction) results in effects which are actually somewhat larger than those in the CBO simulation, particularly in the out years. In the first five years, the DRI model produces an estimated increase in real GNP of $47.4 billion while the CBO/WUMM analysis produces an increase of $35.3 billion. This suggests that the DRI model has a larger response than the WUMM model. Presumably, modeling the smaller cost of capital number (0.5 percent) would lead to smaller effects than those in table 2 if the WUMM model were used.

Over the first six years, the simulations produce an increase of 0.03 percent of GNP (table 2), 0.1 percent of GNP (table 3) and 0.2 percent of GNP (table 4). The numbers in table 3 are similar in magnitude to those of Allen Sinai using the Sinai-Boston Model; Sinai's estimates appear to incorporate a larger effect on the cost of capital, however. ⁶

CONCLUSIONS

The effect of cutting capital gains taxes appears to depend crucially on the effect the tax cut has on the cost of capital. As noted above, the magnitude of the potential effect on the cost of capital is overstated in the CBO simulation and in the comparable simulation presented here in table 4, because of an error in calculating the effect of the capital gains tax. Even if we limit the analysis to the results reported in tables 2 and 3, we see that incorporating the cost of capital adjustment is quite important. In this regard, it is important to note that much of the capital gains tax cut benefits corporate stock. The mechanism through which the cut is translated into a change in the cost of capital is indirect, and probably delayed in impact. To enter the cost of capital effect directly into the investment equation probably overstates the effect on investment, especially in the short run. Indeed, if the mechanism is via an increase in savings, there would be a short run CONTRACTIONARY effect on aggregate demand, which is not captured in any of the econometric analyses. Thus, the results in table 3 are probably overstatements of the stimulative effects of the tax cut.

It is also possible that the capital gains tax cut would have little or no effect on savings or investment, a possibility which is suggested by the literature on savings responses and which is discussed in both the CBO and the CRS reports. Indeed, the capital gains tax cut could reduce the capital stock in the long run if financed by an increase in the deficit.

Although the estimated effect of the capital gains tax cut varies with the modeling of the cost of capital, the effects are negligible and suggest that the capital gains tax cut will have no noticeable impact on the overall performance of the economy.

                               TABLE 2.

² See: U.S. Library of Congress. Congressional Research Service. Can a Capital Gains Tax Cut Pay for Itself? by Jane Gravelle. March 23, 1990. CRS Report 90-161 RCO. p. 15.

³ A brief reprise of the cost of capital measurement dispute may be helpful here. The Council of Economic Advisors initially estimated a cost of capital effect of 3.6 percent if the tax were directly translated into a cost of capital measure. Both the CBO and the CRS studies criticized this measure due to an error in not accounting for the large fraction of capital gains taxes which are never collected because of step-up in basis at death; a correction for this factor would reduce the estimated effect considerably. A similar error was present in the econometric model used by CBO in their econometric simulation. Note that although the percentage change used in the CBO econometric simulation is 1.5 percent, this percentage change refers to the change in the USER cost of capital (the measure incorporated in econometric models, which is inclusive of depreciation) and not the required pre-tax return. Since depreciation is about half of the user cost, a 1.5 percent change in user cost is about a 3 percent change in return.

⁴ Recall that these estimates already incorporate the most important revenue feedback effect, the realizations response.

⁵ The revenues are almost offset in the Table 4 simulation; this simulation is based, however, on an incorrect measure of the cost of capital effect.

⁶ Allen Sinai, Capital Gains and U.S. Economic Performance, The Boston Company, Economic Studies Series, October 2, 1990. This study simulated a reduction in the capital gains tax rate from 28 to 20 percent, roughly a 30 percent rate cut, and found a 0.1 percent effect on real GNP in the first six years.