THAT THE unadjusted balance of the public sector’s financial operations is an unreliable indicator of the stance of fiscal policy has long been recognized by economists.1 Its unreliability stems from its endogeneity with respect to the level of economic activity—that is, the sensitivity of most revenue and some expenditure components to the business cycle. Various indicators of the stance of fiscal policy have tried to purge the balance of its endogenous component; the Fund’s cyclically adjusted deficit is one of several measures used. An additional and much discussed problem with the unadjusted deficit is its failure to take account of the effect of inflation in eroding the real value of the government’s net financial liabilities. In a reflection of both these concerns, the Organization for Economic Cooperation and Development (OECD) regularly publishes an indicator that is adjusted for the influence of both inflation and the business cycle.2
Even inflation and cyclically adjusted indicators have been criticized for their lack of comprehensiveness. Boskin (1988), in his work on the U.S. federal government deficit, has argued that the items now excluded from it can have significant macroeconomic effects, and that certain other items have not been properly measured.3 Buiter (1985), however, has argued that a proper measure of the deficit would take into account all changes in the public sector’s net worth, from whatever source, and has proposed that national and international authorities prepare comprehensive balance sheets of the public sector’s assets and liabilities. This would entail inclusion in the measure of the changes to net worth brought about by, among other causes, an increase in the value of the government’s property or mineral rights (for example, arising from a discovery of natural resource reserves). Furthermore, Buiter has argued that adjusting deficits for the impact of inflation or of the business cycle is not enough to make the conventional figure a useful guide to an assessment of the stance of fiscal policy.4
Kotlikoff (1984, 1986, 1989) has gone one step farther, arguing that even the comprehensive public sector accounting proposed by Buiter would not yield a useful indicator. This second and more fundamental critique takes the view that the basic accounting labels—such as tax or nontax revenue, transfer payments, and interest payments—are economically meaningless.
Kotlikoff proposes what he calls the economic deficit, but this is not a notion that can be summarized by a single number or index. The economic deficit is increased when resources are shifted from younger to older generations. This shift increases aggregate consumption and reduces the rate of capital formation because the marginal propensity to consume of the elderly, who have fewer years left over in which to consume their wealth, is higher than the marginal propensity to consume of the young. Kotlikoff argues that many fiscal policies in the United States (for example, the less than fully funded social security system, or the capital incentives introduced in 1981) would not affect the conventional measures of the deficit but would have massive effects on the economic magnitudes that really matter—the aggregate savings rate and the rate of capital formation.
The purpose of this paper is twofold: first, to give a brief exposition of these critiques of the more conventional indicators of the stance of fiscal policy; second, to assess the usefulness of the alternatives that Buiter and Kotlikoff have proposed. However strong the critiques, the conventional indicators will probably continue to be constructed, but it is still useful to consider exactly how and under what conditions they may be unreliable. If they are often or generally misleading, then the question of whether new indicators can be derived from the alternatives proposed by Buiter and Kotlikoff takes on great importance.
The paper begins with a discussion of how the stance of fiscal policy can be defined and estimated and then proceeds first to present, and then to appraise, the critiques of Buiter and Kotlikoff. Conclusions are presented in the final section. An Appendix compares the cyclically adjusted balance used by the Fund with an alternative based on a simple Keynesian model that allows for differences in the weights placed on expenditure and revenue.
I. Definition and Measurement of the Stance of Fiscal Policy
A statement to the effect that the stance of fiscal policy is conservative is usually supposed to mean that the current set of policies is more restrictive in its effect on the economy than some other set of policies. Thus, fiscal policy stance can only be understood by reference to some norm or base case; to say that a policy implying a budget surplus of X percent when the economy is at a level of activity Z is restrictive requires at least an implicit comparison with some other set of policies—for example, those implying a balanced budget.
Definitional and Conceptual Issues
To apply the concept of the stance of fiscal policy, two questions raised by this definition must be answered. First, how is one set of fiscal policies characterized, and distinguished from another? Second, what is meant by an expansionary effect and its opposite? A difficulty that besets attempts to answer either question is the uncertainty about the period of time over which the two sets of fiscal policy and their effects on the economy are to be compared.
These questions can be approached analytically with the aid of a simple, general model.5 Suppose, for the sake of argument, that an economy can be represented by the following linear system:
where Y, X, Y_1, FP, and OP represent, respectively, the endogenous variables, exogenous variables, lagged endogenous variables, fiscal policy variables, and other policy variables.
The endogenous variables would include the price level, the level of real output, the level of employment, and so on. The fiscal policy variables would include the level of government expenditure in each period— at least, where that level is not partly determined by variations in other economic magnitudes, such as unemployment insurance payments—as well as variables capturing the effect of current tax and benefits legislation. In equation (1) it is assumed that these can be represented as exogenous policy variables in a linear system. In practice, however, they would be built into the structure of the model and would therefore not be so easily isolated. Nonetheless, it may be easier to conceive of them as being separate policy variables whose values can be altered in the same way as other exogenous variables.
To characterize one set of fiscal policies and to compare it with another set, it is necessary to determine two sets of values for FP. However, even determining just one set—the one that characterizes fiscal policy in the base case—poses certain problems. Thus, it is not always obvious what constitutes current policy. For example, under current policy, should government expenditure on goods and services be assumed to be constant in real terms? Or should it grow, and if so at what rate? If such expenditure is thought to be determined in real terms, what mechanism of indexation should be used to convert it to nominal terms? Whatever assumption is made must be somewhat arbitrary, and it remains to be seen whether the conclusions that are drawn from a comparison of fiscal policies would in practice be affected by the assumption that is made.6
Once fiscal policy in the base case and in the alternative case has been characterized, it is also necessary to specify what is meant by an expansionary impact. This is not as straightforward as it might seem. A change in fiscal policy might initially increase real gross domestic product (GDP) above its levels in the base case but subsequently decrease it. An overall effect could be calculated by measuring the present discounted value of the difference in GDP in each period that is the result of the new fiscal policy set. Alternatively, the impact on real GDP in the first year only might be taken into account.
Suppose that we are interested in the impact of fiscal policy on real GDP in the first period. Solving equation (1) for the reduced form of the equation, with real GDP in period 1 (y11) as dependent variable, yields the following:
The indicator of the stance of fiscal policy (IFPS) would then be
where FP2 and FP1 are the new set and the base case set of fiscal policy variables, respectively. In other words, the fiscal policy stance is the product of a series of multipliers and changes in fiscal policy instruments. For example, if the changed instruments are transfer payments to pensioners and defense expenditures, the coefficients in H represent the reduced-form multipliers for these two expenditure categories.
None of the various summary indicators of the fiscal stance, with the exception of the weighted indicator devised by Blinder and Solow (1974), takes this form, so that from a strictly theoretical standpoint neither the Fund’s nor the OECD’s summary indicator appears to be adequate.7 It does not follow, however, that these indicators may not be good proxies for the indicators that would be derived from more complex models.
The difference in form between the Fund’s measure and a model-based measure can be illustrated with the elementary Keynesian model:
where C stands for aggregate consumption, E for autonomous expenditure, G for government expenditure, Y for aggregate income, and T is a simple linear tax function.
It then follows that
with a balanced budget change, where dT = dG, one has
In this case, the indicator of the fiscal stance is simply equal to dG and has the same sign as the change in government expenditure. It is straightforward to show that, even when the purely exogenous component of the change in revenue (dt0) exceeds dG, dY can be positive.
With the indicator of fiscal policy used by the Fund for the World Economic Outlook (WEO) report, the stance of fiscal policy will be unchanged, more expansionary (less contractionary), or less expansionary (more contractionary)—the thrust of fiscal policy will be neutral, expansionary, or contractionary—according to whether the change in government expenditure minus the change in tax revenue from one period to the next equals, exceeds, or falls short of the change in what is called the cyclically neutral balance.8 This condition may be stated as follows:
where FIM stands for fiscal impulse measure, G and T are actual expenditures and tax revenues, Gn and Tn are cyclically neutral expenditures and revenues, and the subscripts 1 and 2 denote time periods.
Cyclically neutral expenditure is defined as gp.Yp, where gp represents the ratio of actual government expenditure to income in a base period in which actual and potential income were deemed to be equal, and Yp represents potential income. Cyclically neutral tax revenues are in turn defined as t.Y, where t represents the ratio of revenues to income in the base period. The fiscal impulse measure can be broken down into its expenditure and revenue components, so that it is possible to speak of the sources of the fiscal impulse. The contribution of the expenditure side (CG) is given by
and that for the revenue side (CR) by
To revert to the notation of the simple Keynesian model, if period 2 is the reference period, then dG is positive when G2 exceeds its previous value by more than the growth in cyclically neutral expenditures, and similarly for dT. If dG = dT, however, then the impulse is zero—the stance of fiscal policy is unchanged. Thus, in the WEO presentation the multiplier effect of revenue and expenditure is the same. It follows that measures of the stance of fiscal policy derived from a Keynesian-type model and the Fund’s fiscal impulse measure could differ not only in magnitude butin sign. The Appendix explores the question of whether this could be a serious problem in practice.9
II. Comprehensive Public Sector Accounting and Indicators of the Fiscal Stance
In two recent papers, Buiter (1983 and 1985) has criticized the conventional measure of the deficit and the fiscal stance indicators constructed by the Fund and the OECD, among others. Instead, Buiter (1983) advocates the use of comprehensive accounting for the public sector, which measures all changes in net worth of the public sector from whatever source.10 He faults the conventional measure of the deficit, even when that measure is expressed in real terms, because it excludes changes to the net worth of the public sector stemming from changes in the real values of the outstanding stocks of public assets. Changes in net worth that are excluded are those resulting from, among others, the impact of inflation on the real value of assets and liabilities whose value is fixed in nominal terms; the depreciation of real assets; exchange rate variations when assets and liabilities are denominated in foreign currencies; changes in relative prices (for example, changes in the value of public mineral rights resulting from a change in the relative price of minerals); and changes in the present value of future tax payments under current tax laws and of future expenditure commitments under current expenditure programs.
In the 1983 paper Buiter does not dismiss out of hand cyclically and inflation-adjusted deficits, noting that “inflation-accounting in the public sector is long overdue” (p. 345). He also argues that a traditional contracyclical fiscal policy can be interpreted as a device to reduce the gap between permanent and current income, a policy that increases economic welfare when consumers are liquidity constrained.
In the later paper, however, Buiter (1985) asserts the need for long-term budgeting that takes into account the future path of revenues and the financibility of the deficit, but he rejects conventional measures of the deficit, whether adjusted or not, on the grounds that none of the simple indicators of the fiscal stance can be derived from a model rich enough in features to be taken seriously.
Buiter recommends that national and international authorities begin constructing comprehensive balance sheets, presumably with the aid of a macroeconomic model, so as to encompass these kinds of changes to net worth in the measure of the deficit. This would be a prodigious but worthwhile task because conventional measures of the public sector balance may seriously misrepresent the options open to a government.
Is this proposal practicable, and, if it is, what purpose would it serve? In particular, could the measure of the change in net worth derived from this estimate replace the conventional indicator, or even the inflation-adjusted and cyclically adjusted version of the conventional indicator?
Unlike the conventional accounting measure, the comprehensive measure would require that assumptions be made about such macroeconomic phenomena as the rates of growth of real and nominal GDP, the evolution of the general price level and of interest rates, as well as the price of assets owned by the government. Changes in net worth from some of the sources noted above would be more easily estimated than others. For just two examples of difficult valuation problems, consider changes in the value of a missile system or in the value of national parks or natural reserves. More tractable would be the valuation of mineral royalty rights, but much uncertainty would attach to the estimate. Changes in net worth resulting from increases in the present value of future expenditures under an entitlement program could be estimated with the aid of a macroeconomic model, but inevitably the assumptions that would need to be made about the model’s exogenous variables and policy instruments would be somewhat arbitrary, as noted in the previous section. In particular, the future course of expenditures under an entitlement program may be determined by current legislation, given the macroeconomic environment, but the same is not true of other expenditures.
Nonetheless, at least some of these changes in net worth could be used to augment the conventional measure. The question remains whether the more comprehensive measure would be a better indicator of the stance of fiscal policy. To illustrate some of the implications of a broader measure, consider the case of a hypothetical economy that enjoys a discovery of a substantial reserve of petroleum in the public domain. The discovery will increase the public sector’s net worth, because it increases the value of the government’s mineral rights. This increase will not be reflected, however, in either the flow of funds deficit or in any of the conventional indicators of the fiscal stance, including those correcting for the effects of inflation and the business cycle. Does the exclusion of changes in the value of such assets from the public sector balance or from the indicators make them misleading?
If the increase is included, then the public sector’s balance is greater, other things being equal, and the stance of fiscal policy, as calculated by the Fund and OECD indicators, is less expansionary (more restrictive) than it otherwise would be. Note that this conclusion does not require that any of the petroleum be extracted. But has there been a contractionary shift in the stance of fiscal policy?
The discovery increases both the public sector’s net worth and the permament income of the economy. Some time can pass before any of the mineral is actually extracted, and unless the economy’s residents treat the discovery as effectively adding to their personal wealth and determine their consumption expenditures on the basis of that wealth, the resource discovery will not lead to an increase in consumption. Nonetheless, the permanent income or wealth of the economy is now higher, and an increase in aggregate consumption is both feasible and possibly desirable.11 An increase in public expenditure or a discretionary reduction in taxes will allow such an increase to take place, so that the application of a simple rule—such as “avoid policies resulting in large positive stimuli to the economy as measured by a conventional indicator of the stance of fiscal policy”—could be quite inappropriate. In such circumstances a large stimulus may be good policy.
Nonetheless, an expansionary fiscal policy will have the same effects on the economy after the discovery as before—some combination of effects such as an increase in domestic absorption and interest rates, and a decline in the external current account balance—and the failure to implement an expansionary fiscal policy will not reduce domestic absorption or lead to a contraction of output. Moreover, inclusion of the change to net worth resulting from the resource discovery in the measure of the deficit and in an indicator of the fiscal stance will mean that this indicator can no longer be used as a guide to the impact of fiscal policy on current macroeconomic activity.
Suppose that instead of a resource discovery, the price of petroleum increases, so that the value of a publicly owned petroleum reserve is increased. Should this increase also be included in the measure of the public sector balance and in indicators of the fiscal stance? In this case, unlike the first case, it cannot be said that the permanent income of the economy is necessarily increased. Such an increase would depend on whether the economy is a net importer of petroleum and on the scale of its net imports in relation to its reserves. To take a specific example, if the increase in the present value of the annual net oil import bill were to exceed the increase in the value of the petroleum reserve, permanent income would undoubtedly fall.12 Thus, although the public sector balance sheet looks better, the private sector’s balance sheet would look worse. In this case it would be the more broadly defined indicator that would give the wrong signal, and a policy based on the rule “avoid large increases in public sector net worth” could give the wrong results.
Some analysts have proposed augmenting the conventional deficit by a measure of the change in the present value of the unfunded liabilities of the social security system.13 At present, both cash-based measures of the government’s financial operations and those based on national accounting treat the operations of the social security system on a cash basis, in the sense that only current operations, not changes in the government’s accrued liabilities, are reflected in either measure. Should these changes be reflected in an indicator of fiscal policy stance?
The answer to this question turns partly on semantics and partly on assumptions about the way the accrued liabilities would affect economic behavior. If the change in the accrued liabilities results from a policy introduced in the current year, it is arguable that any change in economic behavior they bring about should be attributed to fiscal policy. Thus, for example, if one of the fiscal policy variables in the model in Section I represents policies affecting the social security system, and a change in these policies affects current macroeconomic activity, then the policies ought to be included. If, however, the increase in accrued liabilities results from policies taken in the past, it would not be included in the measure of fiscal policy stance determined by the model, unless the assumption was made that increases in the unfunded liabilities in relation to the trend rate of growth of GDP constitute discretionary fiscal policy by definition, analogous to the definition of cyclically neutral expenditure in the Fund’s fiscal impulse measure.
The crucial issue is whether the increase in unfunded liabilities would affect economic behavior. Many economists believe that consumption in the United States has been increased by the failure to adopt a fully funded social security system, although the estimates of the size of the increase vary greatly.14 A case can certainly be made for broadening the conventional measure of the deficit to include this change in net worth of the public sector, but it would be difficult to determine the size of the weight to be attached to it in an indicator of the stance of fiscal policy.
The values of mineral or resource rights and of unfunded entitlement program liabilities are two examples of changes in public sector net worth that are not reflected in conventional measures. Others include contingent claims on government resulting from public insurance programs and contingent liabilities entailed by the possibility of default by beneficiaries of public lending programs, and the decline in net worth entailed by the depreciation of infrastructure and other publicly owned real assets. These phenomena can have significant implications for public policy, but unless their change can be shown to contribute to the impact of government operations on aggregate demand or financial markets, they do not belong in an indicator of the stance of fiscal policy.15
III. Concept of the Economic Deficit
Kotlikoff (1986) has argued that accounting labels such as “taxes,” “expenditure,” “transfers,” and “borrowing” are ill-defined and essentially arbitrary terms with no general basis in economic theory, and that neutral terms like “receipt” and “payment” would be more appropriate. The conventional nomenclature is arbitrary because receipts and payments may be labeled in a number of different ways. Nonetheless, the choice of nomenclature affects the calculated value of the deficit.
Kotlikoff gives an example of a transaction between the government and an individual that comprises a receipt by the government in one year of $1,000 and a payment to the individual of $1,500 ten years later. The receipt in year zero could be labeled “taxes,” and the payment in the tenth year “transfers.” Alternatively, the receipt could be labeled “borrowing,” and the payment “repayment of principal plus interest.” With the first set of labels, the deficit in year zero is $1,000 less than it would be without the transaction, and it is $1,500 more in the tenth year. With the second set, the deficit in year zero is unaffected, and in the tenth year it is increased by the part of the $1,500 that represents interest payments. However, the change in nomenclature obviously has no effect on the behavior of the economy, because the individual’s consumption possibilities are unchanged (Kotlikoff (1986)).
This argument overlooks what appears to be a crucial difference between taxation and government borrowing: taxes must be paid, but generally no one is obliged to lend to the government. Moreover, with the possible exception of social security contributions, the receipt by government of revenue is not paired with an obligation to make a payment at some future date. Finally, even if the classification of various items such as “revenue,” “expenditure,” and “financing” is not clear-cut, classifications are not usually changed from one year to the next.
Taking this last point first, it is an easy matter to demonstrate that with one set of labels a change in fiscal policy can leave the deficit unchanged, whereas with another set of labels the change in policy does alter the deficit. Consider a policy that entails an expansion in the coverage or an increase in the generosity of benefits of a pay-as-you-go old-age state pension program. The expansion of this pension system has no impact on the deficit when the conventional accounting definitions are employed because the expenditures and revenues of the government will increase by equal amounts. However, if contributions are treated as loans to the government and expenditures on pensions are treated as repayment of loans and interest, then the deficit will be affected.
Turning to the first and second points, Kotlikoff’s treatment of social security contributions as loans assumes that such contributions do earn an implicit return, albeit an uncertain one, and also—which may be more controversial—that the obligation to participate in a public pension plan and to make social security contributions does not reduce the wealth of the participant any more than the purchase of a bond would. In effect, the contributions are treated as if they created an obligation on the part of the government to repay the contributors. In the terminology of the Fund’s Manual on Government Finance Statistics, the contributions are a repayable receipt of the government (International Monetary Fund (1986, p. 97)).
The obligatory nature of the transaction is of little economic significance to Kotlikoff. Although many economists would assume that an increase in social security contributions, whether through an increase in rates or an extension in the coverage of the system, would reduce disposable income and hence reduce consumption expenditure, Kotlikoff takes as axiomatic the view that consumption in any period is determined by lifetime income—that is, by wealth. If social security contributions effectively give title to a future income stream of equal present value, they do not reduce wealth. Consequently, they do not reduce the consumption of contributors in any period.
The expansion of a pay-as-you-go system is an instance of a fiscal policy that creates what Kotlikoff would call an economic deficit, since it increases consumption expenditures and reduces the resources available for capital formation. Specifically, when the social security system’s expenditures are increased, the elderly, retired generation benefits from an increase in its wealth; it receives pensions without having to make contributions. Kotlikoff assumes that this generation spends all or most of the increase during its lifetime. Meanwhile, the younger, working generation, which “pays” for the increased expenditure, does not reduce its consumption proportionately, because its wealth is unaffected. As a result, aggregate consumption expenditure increases. In a fully employed economy, the resources available for investment are reduced.
The conventional measure of the deficit through flow of funds accounting has been criticized for not taking into account unfunded liabilities of the social security system, as noted in Section II. Kotlikoff’s critique is more basic, since the creation of an economic deficit does not depend on the need for future increases in rates of contribution to preserve the financial balance of a pay-as-you-go system. Even if rates can remain indefinitely at their initial levels, an increase in consumption and a shift in resources to the elderly take place. Moreover, the assumption that the propensity to consume of the elderly is higher than that of the young is unnecessary. All that is necessary is that perceived wealth increase, and this does happen, given Kotlikoff’s assumption that the contributions of the younger generation are akin to the purchase of securities.
Economic deficits can be created by a reform that replaces one kind of tax with another, even if the yield of the new tax is no different from that of the old. For example, a shift from a tax on income to a tax on consumption would shift resources from the elderly, who are net dissavers— they are in the phase of the life cycle at which their consumption expenditure exceeds their income—to younger generations, who are accumulating wealth and spending less than they earn.16
Changes in the economic deficit in the sense intended by Kotlikoff can even be created by governmental measures that would not be regarded as having anything to do with fiscal policy. An instance would be the introduction of an environmental measure that restricted the use of industrial capital and consequently reduced the value of the existing stock of capital.
How would an across-the-board reduction in personal income taxes be treated in this framework? The answer is not obvious—it would depend, among other things, on how the reduction in taxes was distributed across generations, on the degree of progressivity of the tax system, and on the distribution of income. Above all, it would depend on how the reduction in revenue was to be financed. If the reduction is financed by the issue of bonds now, combined with a subsequent increase in taxes later, the impact of the tax reduction also depends on the distribution of the burden of the future tax increase. When the same people who benefit from the tax reduction pay the subsequent increase in taxes, the reduction in taxes has no effect on the economy. The increase in current income resulting from the reductions leads to an increase in financial saving, and in the private sector’s holdings of public debt of equal amounts. There are no liquidity constraints to be relieved by the increase in current income, and permanent income is unchanged. Hence, there is no increase in consumption.
If older persons benefit from the tax reduction, however, subsequent generations pay for it, and consumption increases because the marginal propensity to consume of the first group, as we have seen, will be higher than that of the second. If the tax reduction is to be financed by the creation of money, the impact of the reduction on consumption will depend on the distribution across generations of holdings of money and other financial assets whose values are fixed in nominal terms. Nonetheless, the use of inflationary financing would have many ramifications beyond its impact on consumption.
The discussion to this point suggests the following conclusions. First, no simple model could determine with any precision the magnitude of the impacts of different measures on consumption and capital accumulation. The most it might do is to indicate whether the impact is positive or negative. Second, the analysis is irrelevant for an understanding of the short-run impacts of fiscal policy in an economy that is experiencing deficient aggregate demand or is significantly cash-constrained; the available evidence on cash constraints in the United States and other economies has to be assessed before Kotlikoff’s critique can be fully evaluated.
The presence of liquidity constraints would change the analysis. Thus, suppose that the taxes paid by a given generation are shifted forward in time from midlife to early adulthood, when current income is typically lower. The life-cycle model would imply that consumption would be high relative to income in the earlier period if households can borrow on the strength of their future income. However, if they cannot borrow, their consumption in this period would be less than it would be if they could borrow.
In these circumstances, a tax reduction now that is financed by an increase in taxes in the later period is akin to a loan, and it is quite possible that consumption will increase. It is important to understand, however, that this increase depends on current income being low relative to future income, and that liquidity-constrained households behaving as the life-cycle model predicts would not invariably increase their consumption by the full amount of the tax reduction.
Recent studies suggest that a significant proportion of consumers in the United States are liquidity constrained, essentially because consumption is more sensitive to fluctuations in current income than would be predicted by the life-cycle model in the absence of cash constraints.17 For example, Hall and Mishkin (1982) found that some 20 percent of consumption is by liquidity-constrained households. Mariger (1986) estimated a figure of 19 percent with his preferred model. Flavin (1981) also found that liquidity constraints are an important determinant of consumption. In view of the relative ease with which consumer credit is obtained in the United States, and the relative absence of institutional rigidities affecting credit markets, it is arguable that liquidity constraints would likely be more important in most other economies, and in particular in less industrialized economies. Rossi (1988), who reported results of tests of consumption behavior for a sample of developing countries, concluded that liquidity constraints are a significant influence. It is also arguable that there is a psychological dimension to the liquidity constraint—some people do not like to borrow, even if “objective” criteria imply they can afford to do so.
It needs to be emphasized, however, that these studies do not point unambiguously to the conclusion that Kotlikoff’s approach is invalid. If 80 percent of consumption in the United States is by households that do not experience liquidity constraints, and these households smooth consumption over time in the manner predicted by the life-cycle hypothesis, tax reductions could be mostly offset by increases in household savings, and vice versa, unless they affect different generations. Altonji and Siow (1987) did not find much evidence against the assumption of perfect capital markets. Moreover, the existence of liquidity constraints taking the form of credit rationing or differential borrowing and lending rates does not automatically rule out the possibility that consumption will not be affected by a tax cut; for example, it is possible that lenders may reduce the maximum value of loans in response to a tax cut, on the grounds that future tax increases will increase the likelihood of default if credit limits are not reduced by an amount proportionate to the increase in current disposable income entailed by the tax cut (Hayashi (1987)).
Another related and problematic assumption underlying the Kotlikoff approach pertains to the interpretation by households of the government’s financing constraint. Would households automatically assume that a tax reduction now would be financed by a tax increase of equal present value at some future date? Any tendency to discount the future at a rate higher than the rate at which the government borrows implies that a tax reduction increases household wealth, and using the reduction to finance consumer durable expenditure—not consumption expenditure in the life-cycle model—would not be irrational behavior.
IV. Appraisal and Conclusions
It is worth emphasizing that any measure of the stance of fiscal policy must be specific to one particular model of the economy. Thus, in “ultra-rational” models in which the unexpected announcement of a future increase in government expenditure raises long-term interest rates and depresses investment and aggregate output in the present, even the sign of a conventional fiscal impulse measure would be wrong (see Buiter (1985, pp. 48–49)).
Kotlikoff’s critique of fiscal indicators based on standard accounting labels is itself dependent on a neoclassical model of the economy and assumes that liquidity constraints are unimportant. A tax reduction now that is financed by an increase later, however, is not a matter of indifference to cash-constrained households contending with imperfect credit markets. In a cash-constrained world, the component of wealth contributed by current income is important, and buying on the installment plan raises the purchase price. In such a world, summary indicators of the budget stance derived from flow of funds accounts can serve a purpose.
Nonetheless, there are some fiscal policy changes—such as a substantial change in the tax regime—whose impact on aggregate demand no simple model could capture and that would not be reflected in a change in the budget balance. The summary indicator approach is probably most reliable for modifications to existing tax regimes and expenditure programs, rather than for wholesale tax reforms and substantial changes in the composition of expenditure. In any case, the summary indicator approach is not intended to capture the impact of fiscal policy on the rate of capital accumulation.
Two other criticisms of the summary indicator approach also deserve comment. It is often argued that the indicator approach is misguided because of its failure to consider the sources of financing of the public sector’s operations. Thus, in the simple monetarist model, fiscal policy cannot be expansionary unless it is validated by an increase in bank financing. Does it follow that it is useless to determine whether the fiscal stance has changed?
This question is partly one of semantics, but there is a point to the exercise regardless of the way an increased deficit is financed. That is because, even in a simple monetarist world with the money supply held constant, an expansionary fiscal policy has some impact on the economy. A tax reduction leading to an increase in the deficit is associated with higher interest rates and capital market pressures. An increase in the deficit entailed by a fortuitous decline in inventory investment is not associated with these effects.18
It is quite true that an expansionary fiscal policy does not result in an expansion of output in all models. Nonetheless, even in ultrarationalist models expansionary fiscal policy is invariably asssociated with capital market pressures and interest rate increases. Whatever the model, the autonomous component of the public sector balance has an effect that differs from its endogenous component. For this reason, the separation of fluctuations in the public sector balances into their autonomous and cyclical components is a valuable exercise, even though the distinctions made are inevitably arbitrary.
A second criticism is that the various summary indicators are easily misused. Thus, a given measure shows that policy became more restrictive at a time when inflation accelerated, and the inference is made that the measure countenances a more expansionary policy, which most people would regard as clearly inappropriate. This inference is incorrect. The estimation of the fiscal stance and its evolution implies nothing about which stance is appropriate. It does not follow that because the fiscal stance has been tightened, it should not be tightened further. The moral to be drawn from this criticism is that summary indicators should be used with care.
Would the WEO indicator be misleading even in a simple Keynesian world? The answer depends on the model’s parameters—the tax rate and the marginal propensity to consume—but it would also depend on the way fiscal policy is implemented.
Fiscal policy can be characterized by using the categories of the fiscal impulse decomposition in Section I, and one of four states must prevail. When the thrust on both sides of the budget is either expansionary or contractionary, the two measures would at least give the same qualitative result; problems arise only when changes are in the opposite direction (see Table 1). How often would changes in opposite directions occur, and how often would the offsetting movements be large enough to change the sign of the impulse?
An examination of the revenue and expenditure impulses calculated using the fiscal impulse method at the central government level for the major industrial countries in the period from 1978 to 1987 shows that revenue and expenditure impulses have been of opposite sign more often than not; specifically, there were 37 instances of opposite sign out of a possible total of 67 (Table 2).19
Nonetheless, in some 29 of these instances of opposite sign, the absolute value of the expenditure impulse equaled or exceeded that of the revenue impulse. For example, in the case of France, where the impulses were of opposite sign in seven years out of ten, the absolute value of the expenditure impulse equaled or exceeded the absolute value of the revenue impulse in six of the seven years. In particular, in 1981 the expenditure impulse was a positive 1.4 percent of GDP, and the revenue impulse a negative 0.2 percent. In these cases, the fiscal impulse measure and the measure derived from the simple Keynesian model would give the same qualitative result.
In the eight remaining instances of opposite sign, the absolute value of the revenue impulse exceeded the absolute value of the expenditure impulse. Here there is a possibility of conflict between the fiscal impulse measure and the simple Keynesian measure, since the expenditure impulse gets a higher weight than the revenue impulse. Thus, in the United Kingdom in 1986, the revenue impulse is a positive 0.9 percent, which offsets the negative expenditure impulse of 0.7 percent. If the weight attached to revenues were half that of expenditure, then the Keynesian measure would be negative. Nonetheless, even in these cases of opposite sign, the difference between the two measures would not typically be very great. In most of these cases the absolute value of the impulses as a percent of GDP is not very large for either revenue or expenditure, and the fiscal impulse as a percentage of GDP is small. The sign of one measure could differ from the other, but neither measure would be large, and in view of the margin of uncertainty attaching to such calculations, the difference would not be significant.
Similar patterns are evident in the expenditure and revenue impulses calculated at the general government level (Table 3). Out of a possible 67 instances, 35 are of opposite sign. However, in 16 of these the absolute value of the expenditure impulse equals or exceeds that of the revenue impulse, so that the fiscal impulse and the simple Keynesian measure would give the same qualitative result. Again, in most of the remaining 19 instances of opposite sign, where the two measures could give contradictory results, the absolute value of both measures is relatively small, so that the difference between the two measures would not in general be significant.
|Revenue Impulse||Expenditure Impulse|
|Expansionary||Same qualitative result||Ambiguous|
|Contractionary||Ambiguous||Same qualitative result|
Nonetheless, there are some instances of opposite sign where the absolute values of the two impulses are large. For example, in Italy in 1982, a positive expenditure impulse of 1.2 percent of GDP is offset by a negative revenue impulse of 2.6 percent of GDP. With no difference in weights, the overall impulse represents a withdrawal of stimulus of 1.4 percent of GDP; but if the revenue impulse has a weight of only half that of expenditure, no withdrawal of stimulus would be estimated.
Thus, the fiscal impulse measure does not often give seriously misleading results in a simple Keynesian world, although it is important that the impulse be disaggregated into its expenditure and revenue components so that different weights can be applied to them if necessary.
What of a slightly more complicated Keynesian model in which each of the different categories of revenue and expenditure has a different multiplier? Would the fiscal impulse measure give a different result than the indicator derived from the reduced form of the model? With a model with many expenditure and tax multipliers, the possibility increases that the fiscal impulse measure and the indicator derived from the model would give results that were substantially different. This is particularly true if both the multipliers of revenue and expenditure measures differ substantially from one another, and if the composition of revenue and expenditure measures differs substantially from one year to the next. For example, the weight that would be derived from the model for transfer payments would be less than the weight derived for expenditure on goods and services. Nonetheless, if the marginal propensity to save of transfer recipients is low, the difference between the weights will not have much practical importance.
Even if the difference between the weights for the various expenditure and revenue categories is significant, the practical importance of distinguishing between them depends on whether or not the expenditure and revenue categories tend to vary together. If they do not, then the fact that an expansionary shift in the stance of fiscal policy might in one year be brought about by a reduction in income tax and an increase in welfare payments, and in another by a reduction in social security contribution rates and an increase in military expenditure, would pose problems for the fiscal impulse measure.
|Germany. Fed. Rep. of|
GNP for Canada, the United Kingdom, and the United States. For definition of central government operations, see International Monetary Fund (1988, pp. 75–76).
GNP for Canada, the United Kingdom, and the United States. For definition of central government operations, see International Monetary Fund (1988, pp. 75–76).
|Germany, Fed. Rep. of|
GNP for Canada, the United Kingdom, and the United States. For definition of general government operations, see International Monetary Fund (1988, pp. 75–76).
GNP for Canada, the United Kingdom, and the United States. For definition of general government operations, see International Monetary Fund (1988, pp. 75–76).
Determining how much of a difference disaggregation would make to the indicator of the fiscal stance would require a fully specified model of the economy, from which the multipliers or weights to be attached to the various categories of expenditure and revenue could be determined—an exercise beyond the scope of this paper. Nonetheless, some insight into the limitations of the aggregated approach on the expenditure side can be gained by a simple experiment that applies a lower weight to transfer payments than the weight applied to the other expenditure categories.
One indicator can be constructed by applying the lower weight separately to the component of transfer payments deemed to represent the discretionary component of fiscal policy, and the higher weight to the sum of the discretionary components of the other expenditure categories. This indicator can then be compared with an indicator constructed by the application of a weighted average of the two weights to the discretionary component of total expenditures. This latter indicator makes no allowance for the possibility that the discretionary component of transfer payments could move in the opposite direction from the discretionary component of the other expenditure categories.
This exercise was carried out for the major industrial countries using data from 1979–86. The discretionary change in each expenditure category was determined by taking the difference between the growth of actual expenditures—less unemployment insurance benefit payments—and the growth of trend expenditures, with the trend value constrained to equal the actual value in 1978.20 Nontransfer expenditures were assigned a weight of 1.0, and transfers a weight of 0.5.
The first and disaggregated indicator can be expressed as
where TR is the increase in transfer payments, G is the increase in other expenditures, and the subscript T stands for trend. The second measure can be expressed as
where w is given by the average ratio of transfer payments to total expenditures for the 1978–86 period.
The ratio IFPS2/IFPS1 is in most years close to unity for the major industrial countries in the 1979–86 period, but there are many instances when the ratio is significantly different. Expressed as a percentage of GDP, however, the differences between the measures in these years are less than or equal to 0.2 percent in 49 instances out of 54. However, in two cases—Japan in 1982 and France in 1979—the difference equals or exceeds 1.0 percent (Table 4).
|Germany, Fed, Rep. of||0.7||0.4||–0.3||–1.3||–0.9||0.5||–0.5||–0.6|
|Germany, Fed. Rep. of||0.76||0.34||1.04||0.94||1.01||1.11||1.21||1.05|
|Germany, Fed. Rep. of||–0.2||–0.3||—||0.1||—||0.1||–0.1||—|
AltonjiJoseph G. and AloysiusSiow“Testing the Response of Consumption to Income Changes with (Noisy) Panel Data,”Quarterly Journal of EconomicsVol. 102 (Issue 2May1987).
BlinderAlan S. and Stephen M.Goldfeld“New Measures of Fiscal and Monetary Policy, 1958–73,”American Economic ReviewVol. 66 (December1976).
BlinderAlan S. and Stephen M.Goldfeld and Robert M.Solow“Analytical Foundations of Fiscal Policy” in The Economics of Public Financeed. byAlan S.Blinder and others (Washington: The Brookings Institution1974).
BoskinMichael J.“Concepts and Measures of Federal Deficits and Debt and Their Impact on Economic Activity” in Economics of Public Debted. byKennethArrow and Michael J.Boskin (New York: St. Martin’s1988).
BossonsJohn and D.P.Dungan“The Government Deficit: Too High or Too Low?”Canadian Tax JournalVol. 33 (January-February1983).
BuiterWillem H.“Measurement of the Public Sector Deficit and Its Implications for Policy Evaluation and Design,”Staff Papers.International Monetary FundVol. 30 (June1983).
BuiterWillem H.“A Guide to Public Sector Deficits,”Economic PolicyVol. 1 (November1985).
ChandSheetal K.“Summary Measures of Fiscal Influence,”Staff PapersInternational Monetary FundVol. 24 (July1977).
EbrillLiam P. and OwenEvans“Ricardian Equivalence and National Savings in the United States,”IMF Working Paper 88/96 (unpublished; Washington: International Monetary Fund1988).
FlavinMarjorie“The Adjustment of Consumption to Changing Expectations About Future Income,”Journal of Political EconomyVol. 89 (October1981).
HallRobert E. and Frederic S.Mishkin“The Sensitivity of Consumption to Transitory Income: Estimates from Panel Data on Households,”EconometricaVol. 50 (January1982).
HayashiFumio“Tests for Liquidity Constraints: A Critical Survey and Some New Observations,”Chapter 13 inAdvances in Econometrics Fifth World CongressVol. 2ed. byTruman F.Bewley (New York: Cambridge University Press1987).
HellerPeter S.Richard D.Haas and Ahsan H.MansurA Review of the Fiscal Impulse MeasureOccasional Paper 44 (Washington: International Monetary Fund1986).
HubbardR. Glen and Kenneth L.Judd“Liquidity Constraints, Fiscal Policy and Consumption,”Brookings Papers on Economic Activity: 1 (1986) The Brookings Institution.
International Monetary FundA Manual on Government Finance Statistics (Washington1986).
International Monetary FundWorld Economic Outlook (WashingtonApril1988).
KotlikoffLawrence J.“Economic Impact of Deficit Financing,”Staff PapersInternational Monetary FundVol. 31 (September1984).
KotlikoffLawrence J.“Deficit Delusion,”The Public InterestNo. 84 (Summer1986).
KotlikoffLawrence J.“From Deficit Delusion to the Fiscal Balance Rule: Looking for an Economically Meaningful Way to Assess Fiscal Policy,”IMF Working Paper 89/50 (Washington: International Monetary Fund1989).
MarigerRandall P.Consumption Behavior and the Effects of Government (Cambridge, Massachusetts: Harvard University Press1986).
Organization for Economic Cooperation and Development (OECD)Economic Outlook (Parisvarious years).
RossiNicola“Government Spending, the Real Interest Rate, and the Behavior of Liquidity-Constrained Consumers in Developing Countries,”Staff PapersInternational Monetary FundVol. 35 (March1988).
SchinasiGarry“International Comparisons of Fiscal Policy: The OECD and the IMF Measures of Fiscal Impulse” (unpublished; Washington: Board of Governors of the Federal Reserve SystemFebruary1986).
Mr. Mackenzie, Senior Economist in the Government Expenditure Analysis Division of the Fiscal Affairs Department, is a graduate of Dalhousie University, Canada, and Oxford University. He thanks Professor Lawrence Kotlikoff for comments on an earlier version of this paper but is solely responsible for the views expressed herein.
Public sector” is used throughout this paper to refer to government in general—that is, both narrow and broader definitions of government.
Boskin (1988) presents a summary of his and other economists’ work on this issue.
See “General Discussion” in Buiter (1985, pp. 68–70).
This exposition is similar to that in Blinder and Goldfeld (1976).
These and other conceptual difficulties in the assessment of the stance of fiscal policy are discussed in Heller, Haas, and Mansur (1986).
Buiter (1985) has made this point. Blinder and Goldfeld (1976) have discussed the same problem with respect to the full-employment surplus indicator. The weighted standardized budget surplus was described in Blinder and Solow (1974). Chand (1977) offered an extensive discussion of summary indicators.
The Fund’s measure and other measures are reviewed in Heller, Haas, and Mansur (1986; see especially pp. 10–11). The presentation here differs somewhat from theirs.
Schinasi (1986) has compared the Fund and OECD measures of fiscal impulse.
As noted in the introduction, Buiter is not the only economist to have argued for a more comprehensive definition of public sector net worth for the purposes of measuring the deficit. However, his proposed definition is particularly broad.
It is feasible to the extent that the economy can borrow externally; it is desirable if current consumption before the discovery was not excessive.
The argument needs to be qualified to take account of the possibilities of substituting other goods for petroleum. The argument supposes that these are not very great.
For example, Bossons and Dungan (1983) made such a suggestion in their study of public sector deficits in Canada.
Boskin (1988) discusses this issue.
Boskin (1988, p. 90) states: “It is unclear whether those who have attempted to generate greater information on government assets and liabilities really believe that a net worth variable is the appropriate one (whether adjusted for inflation or cyclical conditions) to enter as a measure of the government’s economic impact.... It is my opinion that such estimates are useful primarily to provide measures of national wealth and to place concern about government liabilities in better perspective.”
Taking a simple example in which a proportionate tax on income is replaced by a proportionate tax on consumption, the tax rate on consumption has to be greater than the rate of the tax on income to generate the same revenue, since aggregate income exceeds consumption. The elderly must then pay more tax under the new regime because their consumption exceeds their income.
To use the IS-LM framework, in the first case the IS curve shifts to the right against a fixed and vertical LM curve, and interest rates must rise to crowd out interest-sensitive expenditure. In the second case, the IS curve shifts to the left, and the comparative static result is that interest rates decline.
When the value of the expenditure or revenue impulse rounds to zero, the impulses are treated as having the same sign.
Trend expenditures were then assumed to grow at the same rate as nominal potential GDP; that is, real potential GDP plus the rate of change of the GDP deflator. This is the WEO procedure.