Deadweight losses and the saving response to a deficit.

Fremling, Gertrud M. ; Lott, John R., Jr.

DEADWEIGHT LOSSES AND THE SAVING RESPONSE TO A DEFICIT

Several recent empirical studies have challenged the conventional wisdom that deficits cause a higher interest rate. Many economists are, however, critical of the seemingly implied theoretical justification--that every individual fully recognizes the size of the deficit and the extent to which it increases future tax liabilities. This paper demosntrates that far weaker information assumptions are needed to obtain an aggregate savings response equal to the deficit: a reallocation of deadweight losses causes Ricardians to save more than their share of the deficit to keep consumption unaffected, allowing for a substantial fractrion of Keynesians, who save less than their share.

1. INTRODUCTION

The saving behavior in the "Ricardian Equivalence Theorem" is often criticized for relying on perfect foresight; e.g., Buchanan [1976] and Evans [1985,85]. Even David Ricardo himself rejected what is nowadays termed the Ricardian Equivalence Theorem, on grounds that people are not sufficiently aware of future tax liabilities (see O'Driscoll [1977]).

This paper demonstrates that the saving response obtained by the Ricardian Equivalence Theorem--saving oing up by the deficit, with a constant interest rate--can be obtained with weaker rationality assumptions when deadweight losses associated with nonlump-sum taxation are accounted for. Deadweight losses have been overlooked in the deficit debate, possibly because the deadweight loss estimates in the public finance literature were rather small up until a few years ago; the first mention of deadweight losses we have found is the brief theoretical discussion of transactions costs in collecting taxes and issuing debt in Barro [1974,1109]. Deficits and welfare is an issue currently debated, but the question posed is usually the reverse of the one asked here; while we analyze how welfare losses affect the macro consequences of a deficit, recent studies have focused on whether deficits increase or decrease welfare. For instance, both Chan [1983] and Barsky et al, [1986] have studied whether deficits are desirable or not though analyzing how uncertainty of present and future after-tax income are affected.(1) Futher, a recent article by Hansson and Stuart [1987] estimated the losses from labor and capital taxes in a current and a future time period to evaluate the welfare consequences of a reallocation of taxes over time.

The theoretical argument made in this paper might be viewed as an obvious elaboration of the Ricardian Equivalence Theorem. Nevertheless, given current parameter estimates of deadweight losses, the argument turns out to be very important quantitatively. A simple illustration will demonstrate that far from everyone--perhaps less than half of the population--needs to have rational foresight about the deficit in order for a deficit to generate a one-to-one response in saving.

II. THE BEHAVIOR OF "RICARDIAN INDIVIDUALS" The Effect of Taxes Altering Efficiency and Deadweight Losses in Each of the Two Time Periods

The standard Ricardian Equivalence Theorem is arrived at assuming lump-sum taxation, with no deadweight losses accompanying taxation. According to the theorem, a deficit has the equivalent effect of current taxation, because people anticipate the future tax liability incurred with the deficit. These Ricardian individuals have no reason to alter their labor supply, the production techniques, or their consumption pattern. All that a deficit therefore results in is an increase in savings that is exactly equal to the deficit.

How would the analysis change if nonlump sum taxations is allowed for, and associated with deadweight losses? One shoudl have to analyze which particular tax is cut today and which is raised tomorow when the deficit is generated, and how both affect the desired levels of labor suplly, consumption, etc., today and in the future. The main part of the paper focuses on altering the income tax rate and how this shifts deadweight losses over time. Deadweight losses will be assumed to arise when taxes make the economy work less efficiently. Such inefficiencies can be caused by labor and capital going from market to nonmarket production, or from more resources being devoted to tax planning. As resources become more poorly allocated, this can be viewed as the production possibilities curve shifting in.

Lowering taxes, creating a deficit, would then (with constant inputs) increase current output, in correspondence with the smaller distortions. It would also cause Ricardian individuals to anticipate the increased future tax burden and hence the future higher deadweight loss and smaller output. If the present value of the increased future deadweight loss is equal to the decreased deadweight loss today--a plausible first approximation if taxes today and in the future are similar(2)--there would mainly be a reallocation of deadweight losses over time.(3) Assuming only negligible responses in the time path of labor supply and consumption, current agregate demand stays constant, but output rises due to the decrease in distortions. In other words, Ricardian individuals not only save an amount equal to the deficit in anticipation of future higher taxes, but also save an additional amount to offset the change in the distribution of deadweight losses. In such a way they attempt--as in the Ricardian Equivalence Theorem--to insulate themselves from the impact of government finance. However, if everybody were a Ricardian, the individual attempts to keep consumption patterns unaffected could not be realized, since production is now higher. The excess supply of output would put downward pressure on the interest rate. Thus Ricardians would now increase saving by more than the government deficit, and would have to find additional investment, projects with lower returns. If everyone were a perfect-foresighted Ricardian, a "pure" reallocation of deadweight losses would thus imply deficits cause people to save more than predicted by the Ricardian Equivalence Theorem, with lower--not constant--interest rates.

In the above discussion, for simplicity a "pure" reallocation of deadweight losses over time was assumed; and as pointed out, such a reallocation can be caused by inefficiencies arising in production, such as switching from market to nonmarket sectors or from tax collection costs. These are certainly not the only ways a reallocation of deadweight losses can occur. Another example is the misallocation between work and leisure; lower income taxes today should cause a smaller distortion between the two today, and higher income taxes in the future would raise future distortions. However, in the labor-leisure misallocation case, output, consumption, saving and interest rates are not only affected by the reallocation of deadweight losses, but also by the changes in the labor supply directly. (As discussed later, this does not pose any problem, as the other effects will work in the same direction.) As far as any other distortions are involved, the analysis carried out here still applies, as distortions and thus deadweight losses can be presumed to rise in the period that taxes are raised, and to fall in the period that they are lowered. One would then have to consider other possible effects, as with the labor-leisure case. Yet, despite the possibility of other misallocations--such as between labor and leisure--this "pure" reallocation scenario with switching between market and nonmarket sectors is likely to be very important. A large share of the deadweight losses estimated in the literature reflects the high labor supply elasticity for women (see later discussion), which probably is caused by switching between home and market production rather than between leisure and market work.

A simple two-period diagram can be used to illustrate the effect on the interest rate. Figure 1 shows indifference curves for private consumption in the two periods and the consumption possibilities fontier. The frontier reflects (for any given technology, labor input, and initial capital stock) the tradeoff between consumption in the two periods.(4)

If tax rates in period 1 are lowered (increasing efficiency) and raised in period 2 (decreasing efficieny), the consumption possibilities curve can shift as in Figure 1. (If there are wealth effects, a lower or higher indifference curve would be reached.) With public consumption unaltered in the two time periods, there will clearly be more output in period 1, and maximum attainable private consumption in period 1 is therefore higher.(5) The curve will also be flatter.(6) The change in the slope corresponds to a lower interest rate, which is needed to induce more consumption today and less tomorrow.(7)

Substitution and Wealth Effects Across Time Periods

The changes in tax rates would cause a substitution effect in the allocation of labor. The discussion above assumed that the redistribution of taxes towards the future would shift labor from nonmarket to market production in period 1, and from market to nonmarket production in period 2. However, the changed taxation of market effort does not merely cause a substitution between sectors, but also a substitution of labor effort between the time periods. The net-of-tax wage increases in period 1 and decreases in period 2, because of the tax change as well as the concomitant change in efficiency. This clearly results in a substitution effect where labor is reallocated from period 2 to period 1.(8) This substitution can be illustrated in a C1 -- C2 figure. Like the previously discussed change in efficiency, it causes a shift in the consumption possibilities curve like the one already shown in Figure 1, increasing the maximum C1 and lowering the slope.(9,10) Again, the lowered interest rate can be explained by the increased supply of output in period 1 and decreased supply of output in period 2: the interest rate has to decrease to induce a corresponding change in the consumption path. The substitution of labor across time periods would thus reinforce the results of the previous section.

If labor is reallocated over time, the preferred consumption path could also be affected, although it is not clear in which direction. Diagramatically, the indifference curves in the c1-c2 space would be altered. If consumption and leisure were substitutes, the shift in labor from period 2 to period 1 would be accompanied by a shift in consumption preferences towards consumption in period 1. If the higher output, caused by the increased work effort, was coupled with an absolute rise in consumption demand that was even greater than the rise in output, the net effect would actually be a higher interest rate. However, even if consumption and leisure are substitutes (which is not absolutely clear), empirical studies suggest that the increased consumption demand would be smaller than the increased output, and the net effect on the interest rate is still therefore negative.(11)

Finally, there can be wealth effects from altering taxes over time. Such a wealth effect arises when the marginal deadweight losses in the two time periods differ, for instance because the tax rates are different to begin with. To maximize welfare, deficits ought to be used to reallocate taxes from periods with high marginal deadweight losses to periods with low ones, which typically translates into evening out tax rates (See Lucas [1986] for a discussion of "Ramsey's smoothing principle").(12) However, it cannot be assumed that deficits per se will raise wealth. Further, even if it was known how wealth was affected, no conclusions about consumption, income, and the interest rate could be drawn without making some strong assumptions with regards to the preference function.(13)

III. COMBINING RICARDIANS AND KEYNESIANS

Not all decision makers can be expected to behave in a fully informed way. To abstract from the various degrees of rationality, suppose that each individual is either a strict Ricardian or a strict Keynesian. The essential difference with respect to deficits is that Keynesians fail to anticipate the future consequences--that a deficit today will cause a tax increase tomorrow. Keynesians know their present disposable income, including the nonmarket income part. Even if Keynesians have no knowledge of deadweight losses per se, they observe the net-of-tax return from market-sector activities, and rationally allocate their work effort between market and nonmarket efforts within any given time period. Reduced taxes today--meaning a reduced tax wedge between market and nonmarket work--increases the total disposable income of Keynesians today, and the increased disposable income in turn increases consumption and savings, in accordance with the marginal propensities to consume and to save. Thus, while a deficit--lowering taxes today--will will not induce as large of a savings response among Keynesians as among Ricardians, there is still a positive savings response, due to the higher disposable income.14

The issue of this paper is whether or not, in the aggregate, with some well-informed and some less well-informed agents, it is realistic that a deficit can generate a compnesating amount of savings without raising the interest rate. To answer this question, we ask what assumptions would be necessary to make regarding the percentages that are Ricardians and Keynesians. In order to arrive at figures for the percentages, it must be specified mathematically how the two groups behave, and then empirical estimates for the variables involved have to be used. The modeling can fortunately be made quite simple, because the question is what is required to leave the interest rate unaffected; hence, possible interest changes as well as the interest sensitivities are irrelevant and need not be considered in the model.

A deficit, d, in period 1 lowers taxes by the same amount. For simplicity, only considering the effects of increasing efficiency in period 1 and decreasing it in period 2 as described in section II above, disposable income goes up not only by the tax reduction itsel, but (due to reduced deadweight losses in period 1) by an additional amount, dr, where d, again stands for the deficit and r for the marginal deadweight-loss-to-tax ratio.15 The increase in disposable income going to the Ricardians is saved entirely, because--knowing that period 2 disposable income will be reduced--they attempt to keep their consumption path unaffected. The increase in disposable income going to Keynesians is on the other hand consumed: with the traditional Keynesian consumption function, only a fraction, s, of the additional disposable income is saved. If w denotes the share of the deficit (tax reduction) in period 1 that goes to the well-informed Ricardians, the total saving response for the two groups combined, denoted by k, equals k=w(d+dr) + (1-w) (d=dr) s. (1) As equation 1 is based on the assumption that the aggregate saving response is equal o the deficit, without any change in the interest rate it is also true that k=d. (2) Replacing k with d in equation (1), factoring out, eliminating the ds on both sides, and rearranging, yields the following required proportion of the tax cut going to Ricardians: w = (1 - s - sr) / (1 - s - sr + r). (3) As the denominator exceeds the numerator by r, the fraction is obviously less than 100 percent. It is also clear from the formula that the higher the marginal propensity to save by the Keynesians, the lower is the percentage Ricardians necessary. Likewise, the higher the marginal deadweight loss raio, the lower is the proportion that have to be Ricardians. This last conclusion is due to the higher deadweight loss ratio implying a larger increase in current disposable income for a given deficit, hence inducing a larger saving response from both Ricardians (who save their entire increase) and Keynesians (who save a fraction of their increase).

IV. EMPIRICAL ESTIMATES O THE MARGINAL DEADWEIGHT LOSS AND THE REQUIRED PROPORTION OF

RICARDIANS

To arrive at actual numbers for w in equation (3), it is necessary to consider what may be reasonable numbers for the Keynesian marginal propensity to save, s, and the marginal deadweight loss ratio, r. While there is little controversy of s being in the neighborhood of 2, there is great disparity in the estimates of r. We therefore briefly overview the recent public finance literature, and then report alternative figures for w based on alternative estimates of r.

It was long believed that the deadweight loss from taxation was quite small, and it was therefore often abstracted from. Browning [1976] focused attention on the deadweight loss issue, but found relatively modest loss ratios: 8 to 16 percent, counting only the effect of a reduced (market) labor supply. Since then, similar studies have been made, using general, rather than partial, equilibrium analysis. These have found much higher values. Stuart [1984] estimated losses from the labor market of 21 to 57 percent as very plausible, but pointed out that, given recent estimates of labor supply elasticities,16 figures as high as 100 percent cannot be ruled out. Ballard et al. [1985] obtained a range of 17 to 56 percent, taking into account the impact on saving along with the labor market.17 Browning and Browning [1985] indicated losses from labor market distortions of 30 percent and savings distortions of 21 percent from using the current tax system rather than a flat-rate tax.18 Browning [1987] pointed out that improving on his previous partial equilibrium study gives results quite similar to those arrived at through the use of general equilibrium models. In this paper he also emphasized the great sensitivity of the estimates to alterations of the key parameters: using various combinations of reasonable parameter values, he found marginal welfare losses in the labor market ranging from 10 percent up to 303 percent.

From reviewing the different studies, it is evident that a very large portion of the deadweight loss estimates stem from the elastic female labor supply response. Most likely, this great sensitivity reflects household production being a close substitute to market work for women. A large fraction of deadweight losses would then arise from higher tax wedges inducing women to stay home or to work only part-time outside the home, which thus corresponds to the case in section I above.(19) Further, the great variance in the estimates of welfare losses is largely due to the great variance in female labor supply estimates.

Deadweught losses should also include compliance and collection costs. Estimates on these have also changed recently, mainly due to Slemrod and Sorum [1984], who estimated time costs of taxpayer compliance at 5 to 7 percent of revenues collected.(20)

Combining deadweight losses from both the labor market and from compliance and collection costs, a range of 25 to 75 percent does not appear too high. And if recent estimates on labor supply elasticities turn out to be correct, values well over 100 percent are plausible. Because of the great divergence in marginal deadweight-loss ratio estimates, we present a range of figures from 25 to 50 percent. For the Keynesian marginal propensity to save only the figure of .20 is used. (While there is little controversy about this figure, the sceptical reader can easily insert his own numbers into equation (31.) Table I gives the required proposition of Ricardians in the population to yield a sufficient saving response to keep the interest rate constant. As can be seen, even with the lowest estimats of the marginal deadweight loss ration, far from everyone must be a Ricardian. With the high range estimates, a substantial majority is allowed to be Keynesians. To be precise, we should actually talk about the share of income that goes to the two groups: the per capita income is probably relatively higher among the well informed, and therefore the required proportion of the population that must be Ricardian is even smaller than indicated in Table I.(21)

The reader can also use Table I to evaluate whether it seems plausible that a deficit induces a negative or a positive interest response. If the reader's preferred estimate of r generates a w that seems unrealistically high, it implies that the reader feels that, in reality, Ricardians are not plentiful enough to obtain an unchanged interest rate. The saving response (in the absence of interest changes) would be too low, and the interest rate would be forced up. If the w on the other hand appears low, the reader's priors imply the reverse: there are more than enough Ricardians to generate an aggregate saving response equal to the deficit, and a deficit would therefore lower the interest rate.

V. IMPLICATIONS FOR THE EFFECTS OF A DEFICIT

Although many recent studies--e.g., Aschauer [1985], Dwyer [1982; 1985], and Evans [1985; 1987a; 1987b]--empirically have refuted a positive relationship between deficits and real interest rates, serious scepticism has remained as to whether these results are plausible from a theoretical standpoint. The results have been taken to imply that everybody has to recognize future tax burdens, as stated in the Ricardian Equivalence Theorem. This paper points out that results of a zero or negative relationship can be reached with much less stringent information requirements, and must not be rejected on theoretical grounds.

From a macroeconomic point of view, this paper also suggests a somewhat disturbing conclusion--the empirical difficulty in estimating how deficits affect saving. The effect of a deficit can vary not only depending on the degree to which people are informed, but also according to the size of the marginal deadweight-loss ratio. As is apparent from the public finance literature, this deadweight-loss ratio in turn is highly sensitive to the marginal tax rate and to the supply elasticity of labor, which largely reflects the role of women in the labor force. Perhaps the interest response in some countries and during some time periods is positive, in other negative.(22) An implication of this paper is that, ceteris paribus, countries with relatively high marginal deadweight-loss ratios should display relatively high saving responses to deficits. Although this is a testable hypothesis, it is at present virtually impossible to verify empirically, because (as reported in section IV) studies have yielded intervals for deadweight losses that are so extremely wide that differences between countries cannot be discerned.(23) The same problem (to distinguish differences across counties) holds for recent empirical work on the deficit, such as Evans' [1987b] recent cross-country study.(24)

VI. CONCLUSIONS AND IMPLICATIONS

The main contribution of this paper is demonstrating that the informational requirements to obtain a saving response as large as that predicted by the Ricardian Equivalence Theorem are not very stringent. When deficits are generated, rational individuals who attempt to smooth out their consumption paths not only save for the future additional taxes, but also for the future additional deadweight losses created by these crated by these additional taxes. Given the large deadweight losses found in the public finance literature, the additional saving would far exceed the deficit itself if everyone behaved in this manner. To generate additional saving equal to the deficit, a large fraction of the population can therefore be Keynesian individuals who save far less than their share of the deficit.

* Assistant Professor, University of Houston, and Chief Economist, United States Sentencing Commission, Washington, D.C. Helpful comments have been received from Gerald P. Dwyer, Jr. Paul Evans, Milton Friedman, Richard J. Sweeney, and three anonymous referees from this journal. An earlier version of this paper appeared as Hoover Institution Working Paper #E-86-55/9. The views expressed in this paper are the authors', and do not necessarily reflect those of their respective institutions. Any remaining errors are our own.

(1.) See Fremling and Lott [1987] and Richard J. Sweeney [1988, ch. 9, fn. 1] on whether deficits have any advantages in producing insurance.

(2.) Higher tax burdens typically are assumed to cause higher marginal deadweight-loss ratios, and the tax burden as a share of the taxable sector therefore ought to be evened out across time periods. For a recent discussion of Ramsey's smoothing principle, see Lucas [1986, 125].

(3.) Hansson and Stuart [1987] argue that a reallocation of labor taxes from the present to the future is likely to lower total deadweight losses as the marginal deadweight-loss ratio "(marginal cost of public funds") from a tax on current labor is higher than for a tax on future labor, but that the reverse holds for taxation of capital. For income taxes as an aggregate, the two effects to a great extent cancel each other out: "The net effect is relatively small because the income tax is a mix of a labor tax and an asset tax, and postponing a labor tax is advantageous but postponing a capital tax is not" (p. 492). The wealth effect would thus be small.

(4.) The consumption possibilities frontier can be represented by the equation

c2=A2L2 raise to Beta 2(K1 raise to a1L1B1-c1-g1)raise to a2-g2, (1) where c represent real private consumption: g, real government consumption; L, labor; K, capital; subscripts 1 and 2, the time periods; and A, a, and B, the standard parameters in the Cobb-Douglas p roduction function.

(5.) The maximum c1 does not depend on A2. A higher A1 means that, given K1 and L1, more can be produced today. If no capital is carried over to period 2 this clearly increases the possibility to consume today.

(6.) Using the discussion in footnote 4, a greater efficiency in period 1 and a lower one in period 2 can be represented by a higher A1 and a lower A2. Changes in these affect the slope of the consumption possibilities frontier as follows: d squared c2ldc1dA1=-A2L2 raise to Beta 2a2K1 raise to a1L1 raise to Beta 1(a2-1)(K1+A1K1 raise to a

L1 raise to Beta 1-c1-g1)raise to a2 raise to -2>0, (2) d squared c2ldc1dA2=-L2 raise to Beta 2 a2(K1+A1K1 raise to a1L1 raise to Beta1-c1-g1) raise to a2 raise to -1<0.

(3)

Thus, both the increase in A1 and the decrease in A2 make the production possibilities frontier flatter.

(7.) Obviously, by definition,-(1+1)=dC2ldC1, where 1 is the interest rate.

(8.) Among others, Aschauer and Greenwood [1985] use a two-period model, allowing a reallocation of taxes to produce a substitution effect of labor supply across time periods. The substitution effect is, at least from a theoretical point of view, widely recognized, and because it is obvious, we will not elaborate on it.

(9.) The maximum C1, does not depend on L2. A higher L1, give technology a nd K1, implies that more can be produced, and therefore consumed in the period 1.

(10.) d squared c2ldc1dL1=-A2L2 raise to Beta 2 a2A1K1 raise to a1B1L1 raise to B1-1(a2-1)(K1+A1K1 raise to a1L1 raise to B1-c1-g1)raise to a2 raise to -2>0

(4) d squared c2kdc1dL2=-A2B2L2 raise to B2 raise to -1a2(K1+A1K1 raise to a1L1 raise to B1-c1-g1)raise to a2 raise to -1<0

(5)

Both an increase in L1, and a decrease in L2 thus lower the slope.

11. The cross substitution effect reported is positive but less than unity in studies by Gl and Becker [1975], Smith [1977], and Thurow [1969]. Altonji's [1986] point estimate is negative, but is not significantly different from zero.

12. For instance, during wars or severe recessions, when public spending is high, a deficit can be desirable since it reallocates taxes to future periods with lower marginal deadweight losses.

13. We can take the example of a war in footnote 12 above, and assume that the deficit causes higher wealth. In addition to the three effects mentioned earlier (changes in A1 and A2, changes in L1 and L2, and possible changes in relative preferences in c1 and c2 induced by the changes in L1 and L2) there is thus a wealth effect. As the wealth effect means reaching a new, higher indifference curve, little can be said about the relative tradeoffs between current consumption, future consumption, current leisure and future leisure. It is possible (albeit not very plausible) that higher wealth causes a strong shift towards future consumption and leisure, so that labor effort today goes up and current consumption goes down, futher lowering interest rates. This would, however, only occur if consumption and leisure today were iferior goods.

14. As a theoretical curiosity, it is conceivable that Keynesians would end up saving their share of the deficit, and that an economy with only Keynesians (and no Ricardians) could have a saving response to the deficit, with no change in the interest rate. However, this would require the deadweight-loss ration and/or the marginal propsensity to save to be unrealistically hig h.

15. Note and measured disposable income would probably increase by more than the deadweight loss, because of a switch from unmeasured nonmarket activities into measured market activities.

16. Table 2 in Stuart [1984] summarizes seven recent studies, indicating elasticities for women ranging from .91 to 4.83.

17. The low estimate in Ballard's study, .17, is based on a zero aggregate supply elasticity. However, while this elasticity may be appropriate for the male portion of the work force, it seems hard to accept as an aggregate figure, given the large portion of women with high elasticities.

18. Their dollar figures are translated into percentages.

19. The studies are based on the elascities of market labor supply, where the supply curve reflects the opportunity cost of market labor. They do not (and need not) distinguish whether the supply curve reflects the value of nonmarket work or the value of leisure. Therefore, these papers do not tell us whether the deadweight loss in a given period arises from a reallocation of labor from market to nonmarket production or from increasing leisure. However, even if we were to assume that women who do not participate in the labor force simply indulge in leisure, the deadweigh t losses and their reallocation over time would still occur, and the basic argument of this paper still applies. (See also the brief discussion in the first part of section II.)

20. This is an average, and it is not known whether a change in taxation causes a smaller or larger change. On the one hand, given that the taxpayer already fills out a tax form, the marginal cost might be less. On the other hand, a higher tax rate may induce people to spend greater effort on using special exemptions and deductions, and to devote time keeping receipts, which indicates that the marginal loss could be higher.

21. There are two good reasons why Ricardians should tend to have higher per capita incomes than the Keynesians. First, being intelligent and well informed also raises income. Secondly, for whatever reason one has a high income, the payoff for making rationally informed decisions is higher.

22. This problem is further exacerbated by the fact that different types of taxes, when reallocated over time, can have different impacts. The U.S. in particular (but also most other industrialized nations) relies primarily on various taxes on labor and capital income for revenue, whereas low-income developing countries rely heavily on taxes on goods and services (e.g., see World Developement Report 1986, Table 23). The argument of shifting deadweight losses from the present to the future still applies to consumption taxes. A deficit can mean that the tax-wedge between consumption of market versus nonmarket goods and services declines in period 1 and rises in period 2. Since nonmarket goods and services, by definition, are produced by nonmarket capital and labor, it is possible (but not certain) that the magnitude of deadweight losses being reallocated across time in similar to the one for income taxation, where the deadweight loss primarily is due to the tax-wedge between market and nonmarket income. A crucial difference with the reallocation of consumption taxes is, however, the direct on the consumption path, and therefore the effect on saving. If the tax rates on (market) goods is lowered today and raised tomorrow, consumers will raise consumption today at the expense of consumption tomorrow. The incentive to increase first-period consumption therefore counteracts the incentive to save for future increases in taxes and deadweight losses, but without good empirical estimates it is impossible to judge which effect dominates.

23. Hansson and Stuart [1985] show that Sweden, a high tax country, is likely to have high "marginal costs of public funds"--probably between .67 and 4.51. It is difficult to judge whether this is much different from the numbers for the U.S.

24. While his study casts further doubt on the proposition that deficits cause higher interest rates, the confidence intervals for the parameters ar so large that differences between countries cannot be discerned.

REFERENCES Altonji, Joseph G. "International Substitution in Labor Supply: Evidence from Micro Data."

Journal of Political Economuy, June 1986, S176-S215. Aschauer, David Alan. "Fiscal Policy and Aggregate Demand." American Economic Review,

March 1985, 117-27. Achauer, David Alan and Jeremy Greenwood. "Macroeconomic Effects of Fiscal Policy."

Carnegie-Rochester Conference Series on Public Policy, Spring 1985,91--138. Ballard, Charles L., John B. Shovern and John Whalley. "General Equilibrium Computations of

the Marginal Welfare Costs of Taxes in the United States." American Economic Review.

March 1985, 128--38. Barro, Robert J. "Are Government Bonds Net Wealth?" Journal of Political Economy,

November/December 1974, 1095--1117. Barsky, Robert B., N. Gregory Mankiw and Stephen P. Zeldes. "Ricardian Consumers

Keynesian Propensities." American Economic Review, September 1986, 676-91. Browning, Edgar K. "The Marginal Cost of Public Funds." Journal of Political Economy. April

1976, 283-98. ------. "On the Marginal Cost of Taxation." American Economic Review, March 1987, 11-23. Browning, Edgar K. and Jaqueline M. Browning. "Why Not a True flat Tax?" Cato Journal,

Fall 1985, 629-50. Buchanan, James M. "Barro on the Ricardian Equivalence Theorem." Journal of Political

Economy, April 1976, 337-42. Chan, Louis Kuo Chi. "Uncertainty and the Neutrality of Government Financing Policy." Journal

of Monetary Economics, May 1983-3351-72. Dwyer, Gerald P., Jr. "Inflation and Government Deficits." Economic Inquiry, July 1982, 215-29. ----. "Federal Deficits, Interest Rates and Monetary Policy." Journal of Money, Credit and

Banking, November 1985, part 2, 655-81. Evans, Paul. "Do Large Deficits Produce High Interest Rates?" American Economic Review,

March 1985, 68-87. -----. "Interest Rates and Expected Future Deficits in the United States." Journal of Political

Economy, February 1987r, 34-58. ----. "Do Budget Deficits Raise Nominal Interest Rates? Evidence for Six Countries."

Journal of Monetary Economics, September 1987b, 281-300. Fremling, Gertrud M. and John R. Lott, Jr. "Do Deficits Affect the Level of Insurance?" Hoover

Institution Workingg Paper Series, E-87-33, July 1987. Ghez, Gilbert and Gary S. Becker. The Allocation of Time and Goods over the Life Cycle. New

York: Columbia University Press, 1975. Hansson, Ingemar and Charles Stuart. "Tax Revenue and the Marginal Cost of Public Funds in

Sweden." Journal of Public Economics, August 1985, 331-53. ----. "The Welfare Cost of Deficit Finance." Economic Inquiry, July 1987, 479--96. Kormendi, Roger C. "Government Debt, Government Spending, and Private Sector Behavior."

American Economic Review, December 1983, 994-1010. Lucas, Robert E., Jr. "Principles of Fiscal and Monetary Policy." Journal of Monetary Economics,

January 1986, 117-34. O'Driscoll, Gerald P., Jr. "The Ricardian Nonequivalence Theorem," Journal of Political Economy

February 1977, 207-10. Slemrod, Joel and Nikki Sorum. "The Compliance Cost of the U.S. Individual Income Tax

System." National Tax Journal, December 1984, 461-74. Smith, James P. "Family Labor Supply over the Life Cycle." Explorations in Economic Research,

Spring, 1977, 205-76. Stuart, Charles, "Welfare Costs per Dollar of Additional Tax Revenue in the United States"

American Economic Review, June 1984, 352-62. Sweeney, Richard J. "Wealth effects and Monetary Theory," Basil Blackwell, 1988. Thurow, Lester C. "The Optimum Lifetime Distribution of Consumption." Expenditures." American

Economic Review, June 1969, 324-30. World Bank. World Development Report. New York: Oxford University Press, 1986.