NEW CRS STUDY MINIMIZES VAT BENEFITS ON CAPITAL FORMATION.

                       CRS REPORT FOR CONGRESS

A Federal value added tax (VAT) as a possible source of additional tax revenues continues to be discussed in both the academic literature and in the popular press. While arguments advanced for and against such a tax are wide ranging, perhaps the primary argument for a VAT is that in comparison to income taxes it will lead to a higher level of private savings and thus a higher future productivity and a higher standard of living. The VAT is different from income taxes because it is a tax only on consumption and does not tax the income on savings or investment until that income is used for consumption.

This study uses a life cycle, general equilibrium model to estimate the effects of a $60 billion value added tax, comparing the effects to raising the same amount of revenues from an income tax. These results suggest that the effects of a VAT on capital formation and future standards of living as compared to income tax increases are quite modest, raising the capital stock by less than two percent after fifty years, and increasing long run consumption by only a quarter to a third of a percent. These increases are not costless. They occur because society responds to a VAT by initially reducing consumption in order to save.

Such small effects reflect in part the fact that the tax substitution, while important in terms of revenues, is relatively unimportant in affecting a $4 trillion plus economy. The small effects also reflect the difference made by the CHOICE of taxes, which only has an indirect effect on savings, in contrast to a fiscal policy that by directly reducing federal budget deficits directly increases net national savings. In other words, the main positive effect on savings and growth would come from successfully reducing budget deficits using either income taxes or a VAT.

                               CONTENTS

EFFECTS OF A VALUE ADDED TAX ON CAPITAL FORMATION

I. INTRODUCTION

In order to reach the future Gramm-Rudman maximum deficit, the Congress will be faced with either cutting spending programs further, failing to meet the targets altogether, or enacting new taxes, or some combination of these outcomes. A Federal value added tax (VAT) is a possible source of additional tax revenues that continues to be discussed in both the academic literature and in the popular press.

The arguments advanced for and against such a tax are wide ranging. Perhaps the primary argument for a VAT is that as a consumption tax rather than an income tax it will lead to a higher level of private savings and thus a higher stock of capital (i.e., plant, equipment) and a higher standard of living.

There are also other arguments advanced for a VAT. Some believe that the value added tax may be the only politically acceptable source of a significant revenue increase, because legislators may be reluctant to raise income tax rates after the major tax rate reducing reform legislation of 1986. Many economists argue that a consumption tax base is more equitable than an income tax base, or that it produces less economic distortion, although they tend to favor a progressive consumption tax rather than the necessarily proportional VAT. Arguments have also been made, although they are not correct, that the VAT will aid in reducing the trade balance.

There are also a number of criticisms of a VAT. Opponents argue that it is regressive, falling most heavily on the poor; that it will usurp the traditional right to sales taxes as a revenue source of State and local governments: that it will add a new tax with all of the administrative and compliance costs; or that it will be inflationary in the short run. Some critics also fear that it will be a "money machine" -- a relatively painless and hidden means of raising revenue which will lead to an undesirably larger government sector.

In light of these arguments, the effects on capital formation, and the effects of that increased capital stock on future standards of living are of considerable interest. Although there is probably relatively little dispute that a value added tax will lead to increased savings, the magnitude of that increased savings is important. If the effects on capital formation and future productivity are small, this gain may not be worth the sacrifice of other goals such as tax progressivity and simplicity.

This study uses a life cycle, general equilibrium model to estimate the effects of a $60 billion value added tax comparing the effects to raising the same amount of revenues from an income tax. The major focus of the study is on the differentiated effect of a VAT in comparison to income taxes. In more technical terms, this study focuses on the "substitution effect" of a consumption tax, i.e., the extent to which changing the relative price of present and future consumption will induce individuals to consume more in the future and save in order to do so.

The next section of the study explains why a VAT might be expected to increase capital formation more than an equal-yield income tax. Section III explains the basic outlines of the life cycle model. Section IV reports the results of a simulation of a $60 billion VAT. These results suggest that the effects of a VAT on capital formation and future standards of living as compared to income tax increases are quite modest, raising the capital stock by less than two percent and increasing long-run consumption by only a quarter to a third of a percent.

A more detailed exposition of the model is presented in an appendix.

II. HOW A VALUE ADDED TAX INCREASES CAPITAL FORMATION

There are three mechanisms through which a value added tax might increase the capital stock. The first is that simply by virtue of increasing taxes, while holding expenditures constant, the deficit would be reduced. It might, in fact, be the most significant effect. Since such effects on capital formation would be similar for any deficit reduction, these effects cannot be attributed to the value added tax per se, and the effect of deficit reduction per se is not the subject of this paper.

The second mechanism is called the SUBSTITUTION EFFECT. It is one which economists and others who address the savings effect of the VAT emphasize. It is produced because the VAT does not cause a distortion between present and future consumption. Holding income constant, an income tax reduces the return on savings and means that a given amount of savings can finance less future consumption. In other words, the price of future consumption in terms of foregone present consumption rises. This substitution effect by itself means that savings will increase when a consumption tax is substituted for an income tax. A general equilibrium model takes account of the fact that this effect will be partly offset by the resulting fall in the rate of return to capital. As the capital stock increases, the productivity of a new unit of capital (as capital is added to a fixed labor supply) declines. Thus, as savings begins to increase, the effect will drive down the rate of return and this effect will produce a permanent savings response that is less than the initial response.

It is the substitution effect which is the centerpiece for the argument that consumption taxes are preferable to income taxes in inducing capital formation and this substitution effect is the focus of these estimates.

The third mechanism is an INCOME EFFECT. The imposition of a consumption tax as opposed to an income tax will place differential burdens on incomes of different individuals. Essentially, the use of a consumption tax rather than an income tax means that individuals who have consumption levels higher than their incomes will pay more taxes under a consumption tax than individuals whose consumption levels are lower than their income. In the life cycle model used in this study, there is a resulting greater tax burden on the retired population where consumption is occurring partially out of accumulated assets. This life cycle effect also increases savings because retired individuals have a larger propensity to consume simply because they are nearing the end of their lives. This savings effect could be achieved by other mechanisms or taxes besides the VAT, however.

Because the substitution effect on savings arising from a VAT is the primary focus of this study, a version of the model is calculated with tax compensation which will eliminate this direct income effect. This correction allows the calculation of a savings effect without having part of that savings effect rely on taxing the elderly. Such a policy may be undesirable; moreover, there will be an automatic partial compensation in any case because social security and medicare payments are indexed to price changes. Thus, when a VAT causes prices to increase, it will set in motion an increase in the transfers to the elderly which will partially offset their additional tax payments.

III. THE LIFE CYCLE MODEL

In the life cycle model, each individual faces a budget constraint. The present value of lifetime consumption equals the present value of lifetime resources. In this simplified version of the life cycle model, there are no bequests, so that an individual's life time resources at the beginning of his career are simply the present value of his future wages. Over the course of his lifetime, he will save and earn returns on his savings so as to finance consumption in retirement. The amount of savings will depend on the wage rate and rate of return after taxes he faces. In general, when the rate of return after tax rises (because, for example, taxes are removed) individuals will prefer to consume more in later years than in earlier years, and the savings rate will go up.

In an economy, the desire of individuals to save must mesh with the desires of other individuals and businesses to put that savings to work as investment. The level of savings is the sum of all the savings and dissavings of individuals of different ages. The economy is in equilibrium when the demand for assets by savers is just equal to the supply being sold to finance consumption of dissavers plus net investment to maintain the capital stock at a steady state rate of growth. The economy does not adjust and assume a steady growing equilibrium for some time; that is, capital accumulation occurs slowly over a period of many years. During that interim, the savings rate constantly adjusts to new conditions.

Individuals in the economy would, for example, react to a reduction in income taxes (which raises the after-tax rate of return) by saving more. How pronounced this response is depends on how willing individuals are to substitute present for future consumption, and is measured in terms of an inter-temporal substitution elasticity. (An elasticity measures the ratio of the percentage change in quantity to the percentage change in price.)

The additional savings, however, increases the capital stock in the economy relative to the labor supply (which is assumed to be insensitive to the choice of taxes). The consequence of this increase in capital is to drive up the wage rate and lower the pre-tax rate of return on capital. Individuals then adjust their savings to reflect both the changes in the rates of return and wage rates, as well as the tax itself. ¹ Eventually an equilibrium is reached where the wage rate, rate of return, and capital stock reach new steady state values.

This equilibrium depends not only on the preferences of individuals in the economy for consumption over time, but also on the production technology, and how the addition of capital affects the relative wages and rates of return. This effect, in turn, depends on the ability of firms to produce with different ratios of capital to labor, which is measured by a factor substitution elasticity. This elasticity measures the ratio of the percentage change in capital divided by labor to the percentage change in the wage rate divided by the rate of return. The higher the elasticity, the more easily firms can absorb and use increased amounts of capital. (In the extreme case of a zero elasticity, production requiring an irreducible amount of labor for each unit of capital, additional capital could not be absorbed at all and increases in savings could not persist.)

Thus the long-run effect of substituting a VAT for an income tax depends on both the individual preferences for reallocating consumption over time and the ability to produce with different combinations of labor and capital arising from the technological constraints of production.

As noted above savings can be affected by the changes in relative prices of present and future consumption holding incomes constant (the substitution effect) and by changes in incomes. Two alternative models are set up to isolate these effects. In one case, the compensated case, all tax revenues are rebated so that there is no direct redistributional effect of the tax, only a price effect. In the other case revenue yield is fixed, but there are differences in the distribution of taxes across different individuals. (Neither case adjusts, however, for the changes in income deriving from the general equilibrium changes in factor prices.)

IV. ESTIMATED EFFECTS OF A $60 BILLION VALUE ADDED TAX

As explained in section III, it is the substitution effect which is the primary one of interest since the policy question analyzed in this paper is whether IN COMPARISON to an income tax and over the long run a value added tax will substantially increase private savings. In order to isolate the substitution of one tax for another, a VAT is substituted for 15 percent of income tax revenues, so that there is no total change in revenues. Two different models are estimated. In one, the compensated case, the pure substitution effect of this VAT/income tax change is isolated by holding each individual's tax payments constant. ² In the second, the uncompensated case, aggregate tax collections are fixed but tax payments change for different individuals. Essentially, the second, uncompensated, case allows for a shift of the tax burden towards the existing elderly which affects savings rates because of lower propensities to save among the elderly.

The model for calculating these effects is presented in the appendix. It is, of course, a highly stylized model of the economy. Nevertheless models such as these can be used to provide insight into the general magnitude of results we might expect.

As explained in Section II, the key parameters in the model are the inter-temporal substitution elasticity (E) which measures how individuals respond to changes in relative prices in their consumption preferences over time, behavioral measure is the factor substitution elasticity (S) which determines how easily capital can be absorbed in the productive process. If this latter elasticity is low, it is more difficult to use additional capital in production. The combination of these measures of behavioral response and the calibrated values in the model determine the long run effect on capital formation.

Our ultimate interest in additional capital formation, however, is not for its own sake, but rather for the level of sustained additional consumption it will allow in the future. The size of the increase in the capital stock, by itself, does not measure the long run sustainable increase in consumption. First, an expansion of the capital stock only increases one productive factor, and total output will increase by proportionally less than the increase in that factor. Secondly, all of the additional output is not available for consumption, since a fraction of this additional output must be saved to maintain the larger capital stock (which is depreciating) and to maintain growth in the capital stock that is consistent with steady state equilibrium. Thus, the percentage change in long-run sustained consumption is considerably smaller than the long-run percentage change in capital stock.

Such an increase in the capital stock and future consumption is not costless. It is accomplished by initially forgoing consumption in order to increase savings.

Table 1 presents estimates of the effects of a $60 billion VAT rather than an individual income tax increase on long run capital stocks and long run consumption levels. The long run represents the period after the economy adjusts and would probably take in the neighborhood of fifty years to reach a close approximation.

The results both the compensated and uncompensated models are shown. [sic] In the latter case, the effects are larger because the VAT, relative to an income tax, shifts tax burdens to the old. Since the old are paying more, and the young, who have the income to save, pay less, saving is encouraged. A standard unitary elasticity factor substitution case (S = 1) and two variations in this elasticity are shown. The inter-temporal substitution elasticity of .25, which was the preferred estimate by Auerbach and Kotlikoff, as well as a higher substitution elasticity of .5 is used. ³

The results in table 1 show relatively modest effects under any of the assumptions. In the case of most interest, the first row for compensated changes, the capital stock increases by 1.4 percent, and the long run level of consumption increases by only a quarter of one percent. At today's level of consumption, this is equivalent to a consumption gain of about $12 billion, or $60 per person, to occur fifty years into the future. Achieving this gain in the future is not costless. It arises because in response to the tax policy Americans would REDUCE their consumption initially in order to save more.

                               TABLE 1:

In the context of general deficit reduction, these effects are significantly less important than the direct effects on the deficit. For example, if any tax were adopted and used to eventually meet and maintain a zero deficit, the long run effect on the capital stock would be to increase it by 11.6 percent. ⁵ This effect of tax increases via deficit reduction would be several times the size of the effects arising from relative price reductions. Thus, it would appear that progress towards closing the deficit may be much more important for capital formation than how the deficit is actually reduced. In other words, the main avenue available to the government to affect national savings is in its own savings behavior (deficit/surplus policy) rather than the indirect approach of influencing private savings behavior.

APPENDIX

The substitution of value added for income taxes produces a change in savings behavior and a change in the capital stock which leads eventually to a new equilibrium. In this model the labor supply is held constant. The variables in the model are the old and new equilibrium values for output (Q), the capital stock (K), the wage rate (w), the pre-tax rate of return (r), and the level of consumption C. In one case that all taxes are assumed to be rebated in a lump sum fashion. [sic] This approach allows the model to focus on pure substitution effects and is a compensated case. In the other case government tax proceeds are spent as consumption and the real level of government consumption remains fixed. The aggregate level of tax payments is constant, but the tax liabilities shift between the old and the young. This case is the uncompensated case. Individuals make their choices about consumption over time, based on their incomes and preferences for private consumption.

In the case of the compensated taxes, we have the following general functional forms of equations in solving for the five variables: Q, K, r, w, and C (L is the fixed labor supply):

(1) Q = f(K)

(2) r = h(K)

(3) K = m(r,w)

(4) C/wL = n(r,u)

(5) C = Q - I(K)

The first equation is the production function, which relates the level of Q to the level of K. The second equation indicates that the marginal product of capital is a function of the level of capital. The third equation indicates that the level of capital is a function of the rate of return and the wage rate. The fourth equation indicates that the ratio of consumption to wages in the economy depends on the interest rate and the tax rate, u. The last equation indicates that consumption equals total output of the economy less the amount of output devoted to investment (savings), which is in turn dependent on the level of capital.

In the case of a Cobb-Douglas production function, where the factor substitution elasticity is unitary the precise form of equations (1) - (3) is:

(6) [equation omitted]

(7) r = b(Q/K)-d

(8) K = L(b/(1-b))(w/(r+d))

In these equations, b is capital's share of total income in the economy and d is the rate of depreciation of capital. These equations would be somewhat more complex with other production functions, including the standard constant elasticity of substitution (CES) form.

Developing a precise form for equation 4 is somewhat more complicated. If the government rebates all taxes as lump sums then the individual's budget constraint will incorporate the pre-tax incomes, since these amounts are actually received. However, the tax rate will affect the consumption path over time. The relationship between consumption in different time periods, t, assuming a constant elasticity of substitution is:

(9) [equation omitted]

where p is a rate of time preference, c is the level of consumption in the initial working career, at age zero, and E is the inter- temporal substitution elasticity. This relationship indicates that individuals will prefer to consume more in the future, the larger the after tax rate of return, and that the magnitude of that preference is determined by the inter-temporal substitution elasticity, E.

An individual starting a working career faces a budget constraint which requires that the present value of consumption equals the present value of wages. If we assume that productivity in the economy grows at rate g, and the individual lives for T periods and works for T' periods, this budget constraint is:

(10) [equation omitted]

or, solving the integral:

(11) [equation omitted]

Equation (11) provides the relationship between the wage earned by new workers (workers of age "zero") and their initial level of consumption. We must equilibrate, however, total wages and total consumption in the economy. In equilibrium, if the population grows at rate n, the consumption of a worker t years old will be related to the consumption of an individual zero years old by:

(12) [equation omitted]

If we integrate over the consumption of all individuals, total consumption in the economy is:

(13) [equation omitted]

where N is the number of young (age zero) individuals. In this case, zero refers to the beginning of the working life.

To obtain total wages in the economy per young worker, we note that all individuals are earning the same wage, but that there are fewer older workers because the population is growing. If the population is growing at rate n, total wages in the economy is:

(14) [equation omitted]

Without reproducing these equations again, we obtain the ratio of total consumption in the economy to total wages, by dividing (13) by (14), and substituting from (11). In this process, c, w and N cancel out of the equation and we are left with a relationship with C/W on the left hand side and expressions containing r and u on the right hand side, the specific form of equation (4).

The final equation is equation (5). In long run equilibrium, investment will be such as to have the capital stock grow at the growth rate of the economy, n+g, and also replace assets which have depreciated:

(15) C = Q - (n+g+d)K

When we consider an economy where the taxes are not rebated as a lump sum, so that there are both redistributional effects and substitution effects, no changes occur in equations (1) through (3), or their functional forms in (6) through (8). Individuals will now face a budget constraint which incorporates taxes. The discount rate will now be the after tax rate of return, so that whenever r appears alone in equations (10) and (11), r(1-u) will be substituted. Wages in the budget constraint in equation (10) will be wages after tax, or w(1-u). When a value added tax, v, is introduced, as a substitution for part of the income tax, consumption in the left hand side of equation (10) will be multiplied by (1+v) since to obtain a given amount of consumption, the individual must pay (1+v) for each item. (Another way of saying this is that the real wage will fall to 1/(1+v). As a result, equation (11), in addition to substituting r(1-u) for r, will also have the numerator multiplied by (1-u) and the denominator multiplied by (1+v).

We now need another equation to determine how the new levels of u and v will be set so as to yield the same amount of revenue. If the level of government spending is G, then the original tax rate u multiplied by the tax base satisfies:

(16) u(wL+rK) = G

In the new equilibrium, denoting new values with stars:

(17) u*(w*L+r*K*) + vC* = G

If we then set the new level of u*, we can determine the value added tax rate necessary in the new equilibrium to obtain the same level of G.

Finally, equation (15) must be modified to reflect after tax consumption, or:

(18) C = Q - (n+g+d)K - G

These equations complete the model. In order to estimate the effects of a value added tax substitute, we need to first calibrate the model so as to determine an existing equilibrium and then solve the model for the new values.

Some of the data for this model are taken from Dynamic Fiscal Policy, by Alan Auerbach and Laurence Kotlikoff. In particular, we employ their capital output ratio at 3.7, and their pre-tax rate of return on capital of 6.7 percent. All output is measured in units so that the wage rate can be set at l. Their model is modified in several ways, however. The most important modification is the incorporation of depreciation, which is set at 3.2 percent based on the author's estimate for aggregate depreciation of capital in the United States. In addition, we incorporate general technological advance and population growth (both of which are set at a rate of 1.5 percent). Moreover, when calibrating the model for each new value of the inter-temporal substitution elasticity we change the unknown value (the pure rate of time preference) rather than the observed capital stocks and other variables. We use a working life of 40 years and a total lifespan of 55 years (note that these refer to working periods; if individuals begin work at age 20, their life span is 75 years).

² In effect, some of this compensation will automatically occur because transfer payments, including social security, are indexed to price changes. Note that this compensation method corrects for the direct effect of taxes, but not does not correct for the indirect effects of changes in rates of return and wage rates arising from the expansion of the capital stock.

³ Auerbach and Kotlikoff allow labor supply to be variable in their model, although this effect does not appear to be very important. See Alan J. Auerbach and Laurence J. Kotlikoff, Dynamic Fiscal Policy, Cambridge University Press, Cambridge, 1987, Chapter 5.

⁴ These results are for uncompensated case. Auerbach and Kotlikoff consider a compensation scheme which differs slightly from the one used here. They rebate income to individuals who are alive at the time the tax change is introduced so as to keep them at the same level of welfare. Their results consider a full consumption tax/income tax substitution, that is, a total replacement of the individual income tax with a VAT and the effects are, of course, much larger. Using the compensated model to calculate a total VAT/income tax substitution yields results which are very close to their compensated scheme. In the case with unitary substitution elasticities between labor and capital, the compensated model presented in this paper predicts a long run increase in the capital stock of about 13.7 percent while their model predicts a slightly smaller increase of about 12.5 percent. Their results for the uncompensated case are quite a bit larger than predicted by this model; the primary reason for this difference is the incorporation of depreciation which changes the relationships in the production technology significantly. Since the capital stock does depreciate, the incorporation of depreciation into a model of this type provides a more accurate forecast.

⁵ This estimate assumes that the debt stands at 43 percent of output and that the capital stock is 3.7 times output. A $60 billion tax increase would be inadequate initially to close the deficit, but over time, as the debt is reduced and interest payments decreased, it would be more than adequate, assuming no other policy changes and given the projected size of the deficits. These calculations assume that changes in government savings will not alter private savings, however.