CRS Examines Cost of Cutting Capital Gains Taxes

                       CRS REPORT FOR CONGRESS

Capital gains tax cuts have been of longstanding interest and capital gains tax cuts were part of the 1994 House Republican Contract with America. Capital gains tax cuts were included in the 1995 budget resolution that was subsequently vetoed by the President; these provisions have been introduced in the 105th Congress in S. 2. The recent budget agreement allows for a capital gains tax cut.

The permanent revenue losses from cutting the capital gains tax may be larger than those estimated through standard scoring practices for several reasons. The short-run losses are highly dependent on behavioral assumptions. This report explains what factors lie behind the revenue estimates for capital gains tax cuts, which may shift from an annual $12 billion gain to an eventual $20 billion dollar loss in the case of S. 2. (S. 2 contains a 50% exclusion and an elimination of the 28% cap on the tax rate, along with indexing for newly acquired assets -- referred to as prospective indexing).

Three factors influence the revenue estimates of the capital gains tax reductions: the incorporation of a realization response, the phasing in of the effects of prospective indexing and the shift in timing of gain to qualify for indexing.

A fully phased-in, static estimate would be seven times as large as the estimated cost in the first five years of a combined 50% exclusion and indexation of gains on newly acquired assets.

Maintaining the realizations response and considering the long- run, fully phased-in effects, the cost would more than double in the tenth year. Similarly, looking only at the short run, eliminating the realizations response would double the losses in the tenth year.

Finally, in proposals that allow prospective indexing (indexation only for assets purchased after a given effective date) and a mark-to-market provision which allows individuals to declare gain and pay tax, there would be a transitory one time gain in revenue in the short run as individuals realize gain and pay tax to qualify for indexation in the future. This transitory gain, however, would be offset by a loss in the future gain.

Some critics have argued that the feedback effects from current scoring procedures are too small because they do not take into account asset price effects and aggregate growth effects. While asset prices might rise temporarily, there is no reason to expect more than a transitory effect. It is unlikely that any growth effects would be significant in the short run (and they could actually be negative); while growth is possible in the long run, it is of uncertain size and magnitude.

                              CONTENTS

INTRODUCTION

[1] Capital gains tax cuts have been of longstanding interest and capital gains tax cuts were part of the 1994 House Republican Contract with America. Capital gains tax cuts were included in the 1995 budget resolution that was subsequently vetoed by the President; these provisions have been introduced in the 105th Congress in S. 2. The recent budget agreement allows for a capital gains tax cut.

[2] The permanent revenue loss from cutting the capital gains tax may be larger than those estimated through standard scoring practices for several reasons. Moreover, the short run losses are highly dependent on behavioral assumptions. This report explains what factors lie behind the revenue estimates for capital gains tax cuts for individuals. (We do not discuss cuts in the corporate capital gains tax.)

[3] First, estimates of the taxes lost from lowering the capital gains tax rate include an offset for additional taxes collected on increased realizations (gains that result from the sale of assets) that occur because of lower tax rates. Estimates thus depend on the magnitude of this realizations response, which is in turn based on statistical evidence about the relationship between realizations and tax rates. This current offset is substantial relative to the static revenue cost (the cost assuming no change in the amount of assets sold). New evidence on the size of this realizations response suggests that the magnitude used in current revenue estimates may be too large for the intermediate and longer run.

[4] In addition, the long-run cost of the capital gains tax cuts that include prospective indexing or indexing only for inflation after the effective date of the change will be larger because of the growth in the cost of indexing. This effect alone could more than double the cost relative to an exclusion, and the combination of both effects (a smaller realization response and the eventual effect of indexing) could increase the cost of the tax cuts by several times.

[5] Finally, in proposals that allow prospective indexing (indexation only for assets purchased after a given effective date) and a mark-to-market provision which allows individuals to declare gain and pay tax, there will be a transitory one time gain in revenue in the short run as individuals realize gain and pay tax in order to qualify for indexation in the future. This transitory gain will be, however, by a loss in the future gain.

[6] Numerical illustrations are used to gauge the potential magnitude of these scoring effects.

[7] Some critics have alleged that the feedback effects from current scoring procedures are too small because they do not take into account asset price effects and aggregate growth effects. These issues are also discussed.

[8] The revenue estimates in this paper are rough estimates of the magnitude of effects, based on aggregate realizations and average marginal tax rates. The Joint Committee on Taxation (JCT) estimates are the official estimates for budget scoring purposes. Only the estimates in column 2 of table 1 are taken from the JCT.

BASIC REVENUE ESTIMATES

[9] To begin the discussion of revenue consequences, we use the JCT estimates for the capital gains provisions of S. 2, which include prospective indexing, and a 50% capital gains tax exclusion. We then generate new estimates to disaggregate the effects of the exclusion, of prospective indexing, and of the effect of selling (or marking to market) assets in order to qualify for indexation. These costs are shown in table 1.

[10] As we see in table 1, the pattern of revenue loss for capital gains tax cuts in S. 2 is unusual -- it begins with a gain of $12.8 billion and ends with a loss of $20 billion. The loss in the second 5 years ($87 billion) is over three times the net loss in the first 5 years ($26.4 billion). This increase is much larger than the approximately 25% increase that would be expected from nominal growth.

[11] The third column of table 1 estimates roughly the magnitude of the 50% exclusion in isolation. These estimates use the methodology outlined by the JCT in 1990. ¹ They assume a capital gains realizations baseline starting at $205 billion, growing at the nominal rates projected for GDP growth by the CBO. ²

[12] Because this proposal would eliminate the 28% cap on capital gains tax rates, the effective exclusion is estimated at 40%. (While individuals with tax rates of 28% will receive a 50% exclusion, the effective exclusion will be smaller for individuals at higher rates. For example, the 39.6% top rate will be cut in half, to 19.8%, an effective 30% reduction from the current top rate of 28%). ³

      TABLE 1: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS,

R = A e(to the -bt power)

where R is realizations, A is a constant, b is a constant and t is the effective tax rate. The value of b is set at 5.8 in the first two years and 3.5 thereafter. The realizations elasticity (percentage change in realizations divided by the percentage change in tax rate) at any point is -bt. Evaluated at the midpoint of the old and new tax rates (20%), this elasticity is approximately 1.2 in the first two years and 0.7 in the remaining years.

[13] The fourth column of the table adds the cost of prospective indexation of capital gains assuming a holding period of one year. (These estimates assume an effective date starting at the beginning of the first year.) Note that indexation adds very little to the cost in the first year because only one year's worth of indexation is allowed and only for assets bought only a year ago. The next year includes indexation for two years on assets held for two years and one year for assets held a year. The cost of this indexation grows rapidly. ⁴

[14] In the long-run steady state, indexation is the equivalent of a 50% additional exclusion, and thus the total exclusion climbs from 40% to 70% (40% plus 50% of the remaining 60%). This estimate is based on the holding period and on the projections of real growth and inflationary growth in the economy by the CBO. They assume normal steady-state growth in the value of the stock market at the nominal growth rate of the economy. ⁵

[15] The fifth column of the table adds a different form of indexation, one which requires a holding period of three years, as proposed in S. 2. This assumption delays the impact of the indexing provisions for an additional two years.

[16] Note that indexation eventually adds substantially to the revenue loss, almost doubling it after 10 years. The indexation costs will continue to grow, however, and will eventually more than double the revenue loss because of the realizations response. The elasticity of the realizations response falls as the effective tax rate falls and each additional tax reduction is more costly than the previous one for that reason. Only about half of the full effect of indexation is realized by the tenth year.

[17] Finally, the revenue estimates in column 2 reflect another behavioral response: the presumption that individuals will sell or will mark to market (declare gain and pay tax) in order to qualify for indexation. This feature is responsible for part of the initial revenue gain in the first year of S. 2 . This revenue gain will show up as a loss in the future, however, when gains which would otherwise have been realized fail to materialize.

STATIC ESTIMATES

[18] We have discussed three factors that influence the revenue estimates of the capital gains tax reductions: the incorporation of a realization response, the phasing in of the effects of prospective indexing and the shift in timing of gain to qualify for indexing. One way to calculate the magnitude of these effects is to estimate the long-run steady-state, static loss: the loss that would have occurred had these provisions already been in place and under the assumption of no realizations response. These estimates are shown in table 2. They indicate that the total package would cost $185.7 billion in the first 5 years -- seven times the estimates in table 1 -- and $233.9 billion in the second 5 years -- almost three times the estimates in table 1.

      TABLE 2: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS,

[20] Note also that even if the tax cut were restricted to the 50% exclusion, the static estimates are much higher than the estimates in table 1 -- in the tenth year, the revenue cost is almost three times as large.

LONG-RUN ESTIMATES WITH JCT REALIZATIONS RESPONSE

[21] To further explore the importance of the factors underlying the estimates we use the JCT elasticity assumptions to see the consequences of the phase-in and timing effects. Table 3 estimates these long run costs -- these are the costs we might eventually expect (at income levels appropriate to the next 10 years) assuming the JCT realizations response is correct. Note that the costs of the exclusion (except for the first two years) is the same as table 1, but the cost including indexation is much larger. Even compared to the tenth year, the cost has increased by 50%. Thus, if we wish to consider the eventual cost of both an exclusion and indexation, we would need to increase the cost even in the last period of the budget estimating horizon by 50%. In year three, the cost would be over twice as large.

 TABLE 3: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS, LONG-RUN

[23] Thus, even if the JCT realizations response is correct, the long run cost of an exclusion plus indexation would be much larger than indicated by estimates in the budget horizon of 10 years.

THE REALIZATIONS RESPONSE

[24] The estimates of the capital gains revenue loss would be considerably different with different realizations responses. To illustrate the effects of the realization response in the short run, Table 4 presents the static short run estimates assuming no realizations response. In the third year, the cost of the package is roughly twice as large as in table 1; in the tenth year, the cost of the exclusion is almost three times as large and the cost of the combined exclusion and indexing package is twice as large. Clearly, the realizations response is crucial.

[25] There has been considerable disagreement in the past few years over the revenue consequences of a capital gains tax cut. Both the Administration and the Joint Committee on Taxation include in their revenue estimates certain behavioral responses. The most important of these responses, by far, is the expectation that individuals will respond to lower capital gains taxes by increasing realizations of capital gains. These increased realizations of capital gains will then produce additional taxable gains and additional revenue which will offset the static revenue loss. (The static revenue loss is the loss arising from the lower tax rate assuming there is no behavioral change.)

[26] There has been a substantial body of empirical research on the realizations response, which has yielded a wide range of estimates. In particular, studies that estimated the relationships between realizations and tax rates across different taxpayers (micro- data studies) often yielded extremely large responses.

      TABLE 4: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS,

[28] In recent years, new evidence has been presented that suggests that this realizations response may be smaller than that assumed in the past, especially after the first few years. Because of the wide variation in estimates based on statistical analysis, an alternative method of assessing the likely size of the realizations response was prepared. ⁷ This analysis is based on a relatively simple observation -- in the long run, realizations cannot exceed accruals. That is, realizations would equal accruals over a long period of time (year after year) only if individuals sold all assets after holding them less than a year. Indeed, one would never expect that all gains would be constantly realized, even in the absence of taxes and other transaction costs. The observation of historical ratios of realizations to accruals can be used to measure the upper limit of the realizations response and to suggest a likely size of that response. This analysis suggested a lower, perhaps much lower, permanent realizations response than that measured in most statistical studies. Basically, this type of approach provides a "reality check" on statistical estimates.

[29] This analysis suggested that the very large realizations responses found in most microdata studies lead to implausible estimates of changes in realizations responses -- results that are far outside the bounds of historical experiences and far in excess of accrued gains for tax revisions such as those now being considered.

[30] Most critics believed that a major problem with these studies was that they could not control for timing effects. It is advantageous for individuals whose tax rates fluctuate from one year to the next to realize gains in years when tax rates are low. Thus, the relationships found between low tax rates and high realizations could be reflecting in part, or perhaps primarily, responses to temporary changes in tax rates -- responses that would not hold up for a permanent tax rate change. Indeed, the surge in realizations in 1986 when tax rates were scheduled to go up the next year is evidence of the power of this timing effect.

[31] A recent statistical study which used a new approach -- variation in tax rates across states -- to control for transitory effects that had long plagued microdata statistical studies also found much smaller realizations responses. ⁸ These results were consistent with the effects that were suggested by the study of the historical measures of realizations and accruals. If the findings in these recent studies are correct, the revenue estimates for the 50% exclusion could be more than twice as large as they would be based on current revenue estimating assumptions.

[32] Table 5 estimates the long-run effects assuming a lower realizations elasticity suggested by this new research. In calculating these estimates the constant in the realizations formula is set at 1, for an elasticity of 0.2. While the costs are not as large as in the static case in table 2, they are much larger than either the short run estimates in table 1 or the long-run estimates assuming the JCT realizations response in table 3. In the tenth year, the cost of the exclusion has more than doubled, and the costs of the entire package have increased substantially whether one considers the short run or the long run.

[33] It is much more difficult to determine what the realizations response would be in the intermediate term. The original JCT estimates were termed long run but the times series evidence on which they are based is some combination of short-run, intermediate and long-run response. Table 6 estimates the short run costs, assuming after the first two years that the long-run elasticity has fallen to 0.2, illustrating the importance of the choice of long-run realizations response. If the realizations response has settled down to its steady state after a few years, these estimates might be more appropriate.

      TABLE 5: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS,

[34] While the realizations response is the most widely studied behavioral response associated with capital gains realizations, there are other responses. Some could provide negative feedback (e.g., companies might decrease their dividends in favor of retained earnings or share repurchase). There is also the possibility of tax arbitrage -- borrowing to convert income into capital gains, which could result in tax reductions.

 TABLE 6: REVENUE GAIN OR LOSS FROM CAPITAL GAINS REVISIONS, SHORT-RUN

[36] There are two other effects that are interrelated macroeconomic effects. They include asset price effects and aggregate investment effects. These are short run effects. There are also long- run effects that arise from savings behavior.

SHORT-RUN GROWTH AND ASSET PRICE EFFECTS

[37] Some recent studies have emphasized additional revenue feedback from higher asset prices and higher short run output. ⁹ These effects are intertwined since it is via an asset price effect that short-run positive output effects have been estimated; and feedback is argued to occur from both the asset price effect, which increases gains directly, and the short run growth effect which increases aggregate capital income.

[38] The macroeconomic models are models of aggregate demand; they are not specifically designed to model the nuances of tax changes and in general do not have a direct way to incorporate a capital gains tax cut.

[39] It is helpful to begin by pointing out that in a Keynesian-style aggregate demand model there is a difference between a tax cut on the savings side and a tax cut on the investment side. A tax cut on the individual side will add to aggregate demand in the short run if spent on consumption, but not if saved. A tax cut on the investment side (to the firm) will add to aggregate demand if it is used to increase investment, and perhaps if it is paid out to stockholders as dividends and they spend it. But the crucial difference is that translating a tax cut into savings/investment depends on what type of tax cut it is: one to the saver or one to the investor.

[40] It is also to be expected that the transmission process between savings and investment for the capital gains tax cut (for corporate stock) will be indirect. Individuals first have to increase their purchases of corporate stock, which will presumably lead firms to issue more stock (or to repurchase less than they had originally planned). Firms will have more equity to invest, and also during the adjustment period, asset prices in the stock market could increase.

[41] The process by which this occurs will determine any asset price and output effects, and that will depend, in part, on savings behavior. The following discussion outlines how such modeling should be done in theory (although this type of modeling has not been used in the macro models).

[42] First, suppose the savings rate is zero. The individual tax cut will stimulate consumption in the same way as any other tax cut (with the effects dependent on how close the economy is to full employment and the accompanying monetary policy). Individuals will want to purchase more stock and will withdraw funds from other assets, presumably largely from bonds, to do so. That will have the effect of driving up the interest rate, which will make investment financed by debt more costly, but also of driving up the stock market price and making equity cheaper. Firms should reduce their borrowing and issue more stock (or repurchase fewer shares).

[43] The equilibrium solution to this problem is that the share of debt finance should fall, the interest rate should rise, the aggregate cost of capital in the economy should be unchanged, and asset prices should not change. (This is exactly what one would expect if there were no effect on savings). To make this more concrete, one can think of an asset's value as the ratio of net after tax earnings on equity to the after tax interest rate plus a risk premium:

      A = E(1-t*) / i(1-t) + p

[44] While stock values fluctuate in the short run, they should reflect the underlying value of assets. If we think of the earnings as an earnings rate (per dollar of capital invested reduced by outstanding debt), then the asset should be worth a dollar. (We abstract here from those tax rules that may induce a deviation from the cost of assets, such as an investment tax credit). If the asset price rises above that value firms can make an extra profit by issuing stock and buying assets with the proceeds. Thus, there is a market mechanism that returns the asset price to equilibrium, and that mechanism could operate very quickly.

[45] In addition to issuing new shares of stock, another equilibrium effect in resetting the asset value to its original value is the rise in the interest rate in the denominator that offsets the tax reduction in the numerator. The equity return (before tax) will also fall because there is a larger amount of equity sharing profits, and because the equity return is initially larger than the interest rate. The risk premium should also fall because of the lower debt/equity ratio.

[46] Note the order in which these events begin: individuals take money out of the bond market (driving up the interest rate and making the cost of capital larger) and put the money in the stock market, driving up asset prices. Firms reduce borrowing and issue shares, causing the interest rate to fall back a bit and the asset price to fall. This continues until an equilibrium is reached, where the equity return is again equal to the after-tax interest rate plus risk premium. There is no reason to expect that the cost of capital will move down (indeed, it might first rise because of the interest rate effects), but it should not be affected in more than a transitory fashion. At the same time, firms may move quickly to issue shares, and the asset price effect would be fleeting.

[47] This modeling is different from that done in at least some of the macro models, as in the DRI study. There is no set of supply and demand relationships for determining the effect of a something like a capital gains tax cut. Tax changes on the investment side, such as the corporate tax rate and changes in depreciation, can be modeled directly, but there is no mechanism for modeling a capital gains tax cut (or a tax cut on dividends or a change in the individual income tax rate in general).

[48] However, the modeling involves an exogenous rise in the asset price causing a fall in the cost of capital. This approach converts the capital gains tax cut into an investment incentive without having to deal with its transmission as a savings incentive. This makes it a more powerful stimulus because, unlike the investment incentive, it also adds to consumption by individuals and unlike a savings incentive, it adds directly to investment without first contracting aggregate demand. This is like doubling the stimulus. Moreover, the additional asset price rise causes an additional collection of revenue.

[49] None of this is consistent with fundamental supply and demand analysis. Asset prices could rise if savings increased and if there is a lag in adjustment by firms, but that savings effect would also be contractionary in the short run. In any case, this is a temporary price rise, and any additional capital gains arising from it would be offset by smaller gains (or losses) in the future as the asset price returned to equilibrium.

AGGREGATE LONG-RUN GROWTH EFFECTS

[50] A second potential source of growth is an increase in savings as a result of the capital gains tax. The evidence on the savings elasticity does not, however, support much of an effect.

[51] It may be surprising to some to learn that we cannot be sure whether cutting taxes on capital income will increase savings. Economists have long recognized that the response of saving to the rate of return is uncertain due to the opposing forces of "income" and "substitution" effects. When the rate of return rises, a substitution effect might cause an individual to prefer more consumption in the future (because the price of future consumption has fallen in terms of foregone present consumption) and increase savings. At the same time, there is an income effect -- the higher rate of return can allow savings to be smaller and still increase consumption in the future (and in the present as well). For example, if an individual were saving a certain amount for retirement, he could obtain that objective with a smaller amount of savings when the rate of return goes up.

[52] Because of this theoretical ambiguity, it is necessary to turn to empirical research to determine whether private savings will increase, and empirical evidence would be necessary in any case to determine the magnitude of any effect. While it is very difficult to perform this analysis, this body of research suggests that effects of higher rates of return on savings have small positive effects on savings behavior and, in some studies, negative effects. ¹⁰ That is, it is possible that cutting capital gains taxes will reduce savings. Indeed, many life-cycle models of the economy would readily predict a contraction of the economy from a tax increase. ¹¹

[53] Some models obtain large savings effect from capital inflows from abroad. The capital gains tax cut, however, does not affect foreign investors and there is no reason to expect an inflow of capital.

[54] The process of altering the capital stock through a change in the savings rate is a very slow one that takes many years. Even with a large percentage increase in savings, the effect on the capital stock and on economic output will be modest in the short run because savings is very small relative to the capital stock. Effects could appear in the long run, but the magnitude and direction are uncertain and likely small.

CONCLUSION

[55] While uncertainty remains about these estimates, the illustrative calculations in this report show that the eventual cost of a capital gains exclusion could be much larger than suggested by the current revenue estimates. Part of this effect arises from the phasing in of the effects of indexing. The static, long run estimates reflect the value of these tax cuts to taxpayers better than the short run estimates prepared for budgetary scoring purposes. This analysis also suggests that the impact on the budget could be much larger than suggested by current estimates in the long run, and also in the 10-year budget horizon if the large realization response assumed by the JCT does not materialize.

² Congressional Budget Office, The Economic and Budget Outlook, Fiscal years 1998-2007, Washington, U.S. Govt. Print. Off., January 1997.

³ This calculation assumes an average marginal tax rate of 25% according to the JCT's 1990 study (Explanation of Methodology used to Estimate Proposals Affecting the Taxation of Income from Capital Gains), and an additional 5 percentage points if the cap were not in place based on JCT tax expenditure estimates (Joint Committee on Taxation, Estimates of Federal Tax Expenditures for Fiscal Years 1997-2001, Committee Print, 104th Congress, 2nd session, November 26, 1996) for an overall rate of 30%. Cutting that rate in half produces an average marginal tax rate of 15%, for a 40% reduction).

⁴ The pattern of holding periods is based on data in Holek, Charles C. and John P. Shelton, 1989. 1985 Sales of Capital Assets. Draft Paper for Presentation at the 150th Meeting of the American Statistical Association. August 6-10.

⁵ There are some possible adjustments to this estimate that reflect the average inflation share in opposite ways. First, the inflation share would be larger because individuals have an incentive to sell assets that have small gains and would further have an incentive to sell assets with small real gains. On the other hand, there are restrictions that prevent using this provision to produce a loss in S. 2, that would presumably be reflected in an indexing provision. Also, the share of inflation in the nominal gain of assets outside the stock market would be different.

⁶ These issues were discussed in a variety of different articles at that time. See, for example, Gerald E. Auten, Leonard E. Burman and William C. Randolph, Estimation and Interpretation of Capital Gains Realizations Behavior, National Tax Journal, September 1989, pp. 353-374; U.S. Library of Congress Congressional Research Service, Can a Capital Gains Tax Pay for Itself? CRS Report 90-161, by Jane G. Gravelle, March 23, 1990; Gerald E. Auten and Joseph Cordes, Cutting Capital Gains Taxes, Journal of Economic Perspectives, Winter, 1991, pp. 181-192.

⁷ See U.S. Library of Congress, Congressional Research Service, Limits to Capital Gains Feedback Effects, CRS Report 91-259, by Jane G. Gravelle, March 15, 1991.

⁸ Burman, Leonard E. and William C. Randolph, Measuring the Permanent Responses to Capital Gains Tax Cuts in Panel Data, American Economic Review, September, 1994, pp. 794-809. More recent surveys of the literature cover this more recent research; see George R. Zodrow, Economic Analyses of Capital Gains Taxation: Realizations, Revenues, Efficiency and Equity, In Tax Law Review, Vol. 48, No. 3, 1993, pp. 419-527; Jane G. Gravelle, The Economic Effects of Taxing Capital Income, Cambridge, MIT Press, 1994, pp. 143-151.

⁹ See, for example, testimony of David Wyss, DRI/McGraw-Hill, Impact of Capital Gains Tax Reduction on Revenues and the Economy, before the House Ways and Means Committee, March 19, 1997.

¹⁰ For a summary of this literature, see Jane G. Gravelle, The Economic Effects of Taxing Capital Income, Cambridge, MA, MIT Press, 1994, p. 27.

¹¹ Several growth studies prepared in 1990 are discussed in Congressional Budget Office, Effects of Lower Capital gains Taxes on Economic Growth, August 1990. These include a life cycle simulation of the capital gains tax cut.