1. The line

There are two criteria for distinguishing between revenue/expenditure on the one hand, and financing on the other: the “government debt criterion” and the “public policy criterion.”

b. The public policy criterion

Alternatively, transactions are deficit-determining and classified as revenue or expenditure instead of financing, when they further public policy rather than forming part of public sector liquidity management. Unlike liquidity management, public policy-motivated transactions change the prices facing the rest of the economy compared to what they would be if markets were left undisturbed.

This criterion is also imperfect because, in practice, government does not approach financial markets on the same terms as other borrowers. Typically, government can borrow on more favorable terms, for instance, by imposing restrictions on the placement of public institutions’ funds. Moreover, governments often have a policy agenda underlying their ranking of financing sources (the central bank, commercial banks, different private sector groups, foreign sources), which may make them depart from least-cost borrowing/pure liquidity management. In other words, even through its financial intermediation, government may tax, subsidize, or effectively regulate parts of the economy and, therefore, the public policy criterion provides only a blurred analytical distinction between what belongs to above or below the line.

The two criteria for drawing the line generate the same classification of most transactions; however, they diverge for three types of transactions which have led, in practice, to important discrepancies in the estimated size of the conventional deficit. These are: (i) budgetary “net lending;” (ii) external grants; and (iii) debt service.

(1) “Net lending”

Unlike the other budgetary outlays, government lending operations to the private sector involve liability management (overt and contractual), and hence, by the government debt criterion, should go below the line. However, a significant portion of budgetary “lending” is composed of direct capital infusion and of government credit programs undertaken for policy purposes, viz., to supply funds to preferred sectors who would otherwise not have access to financial markets or who would have to pay steeper rates. Given its implicit subsidy element, and the higher than market probability that some of the loans will never be repaid, net lending cannot be defined as pure financial intermediation, and the public policy criterion, then, would classify net lending as part of government expenditure—above the line. ^6/ As illustrated in Table 1 by the case of Venezuela (where the government allocates its petroleum revenues through domestic and external lending programs), the difference in classification can turn a deficit into a large surplus. ^7/

Table 1.

Venezuela: Consolidated Central Government ^1/

(In billions of bolivares)

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/In this and following tables a minus sign indicates a deficit.

Table 1.

Venezuela: Consolidated Central Government ^1/

(In billions of bolivares)

	1985	1986
Balance excluding net lending	32	4
Balance including net lending	24	-9

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/In this and following tables a minus sign indicates a deficit.

View Table

Table 1.

Venezuela: Consolidated Central Government ^1/

(In billions of bolivares)

	1985	1986
Balance excluding net lending	32	4
Balance including net lending	24	-9

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/In this and following tables a minus sign indicates a deficit.

From an analytical viewpoint, neither treatment is completely correct. Unless budgetary loans are uncollectable from the start (which would imply an outright transfer), they contain both pure loan and pure grant components; only the latter should be considered as a public policy element and included as part of the deficit. ^8/ Moreover, the subsidy (grant) component is usually spread over the entire lifetime of the loan, going beyond the budgetary year in which the loan is extended. ^9/

(2) External grants

Since grant aid from abroad represents financing without liability, the government debt criterion would include it with other government revenues. However, by the public policy criterion, grants are added to other foreign financing—below the line—on the argument that no government policy decision can elicit these grants, and, therefore, that the current expenditure that they finance could not take place if the grants are not forthcoming (Chelliah (1973), p. 749). ^10/ Grants are discretionary financing by donors and can vary significantly from year to year. Their inclusion as regular revenue has been said to give an inappropriate confidence in their permanence, though they may have to be replaced by government borrowing at any time. Particularly in developing countries where domestic incomes are very low, the classification of grants below the line can widen the deficit by more than 5 percentage points of GDP (see Table 2).

Table 2.

The Gambia: Consolidated Central Government

(In percent of GDP)

Source: IMF: Government Finance Statistics Yearbook, 1988.

Table 2.

The Gambia: Consolidated Central Government

(In percent of GDP)

	1981	1982
Deficit including grants	-12	-7
Deficit excluding grants	-17	-17

Source: IMF: Government Finance Statistics Yearbook, 1988.

View Table

Table 2.

The Gambia: Consolidated Central Government

(In percent of GDP)

	1981	1982
Deficit including grants	-12	-7
Deficit excluding grants	-17	-17

Source: IMF: Government Finance Statistics Yearbook, 1988.

(3) Debt service

In some countries, it may be argued that present levels of public debt are not sustainable, and that amortized debt may not be voluntarily reinvested in new government bonds. In such cases, replacement financing for amortization could require a policy effort on the part of government akin to that of generating extra tax revenue. Under this scenario, the public policy criterion would suggest the inclusion of amortization above the line, and the resulting deficit would correspond to the government’s gross borrowing requirement, rather than to its net increase in liabilities.

Such differences in classification as described above can substantially affect the measured deficit. Typically, classifications have evolved apolitically, and countries have maintained one consistent treatment of government transactions over time. However, it is clear that much scope exists for distorting the picture through judicious reclassification. Kotlikoff (1988, 1989) makes this point, showing how the government, by relabeling its transactions (as taxes/borrowing, in various combinations with expenditure/amortization) can shift operations from above to below the line (and vice versa) but, essentially, carry out the same policy while choosing to report either a balanced budget, a deficit, or a surplus. If governments do not maintain what might be termed “ethics in accounting standards,” the fiscal deficit ceases to convey useful information.

2. The cash and accrual deficits

The other main conceptual variation among conventional deficit measures is the choice between cash and accrual accounting. At one end of the spectrum is the completely cash deficit, where only government outlays for which cash has been disbursed during the 365-day period, and only actual cash revenues received, are included in the budget balance. At the other end of the spectrum is the completely accrual deficit, which attempts to capture the actual net resource pre-emption of government—the consequences of its policy decisions—during the fiscal year, regardless of whether or not transactions have actually been paid for. Thus—an important example—depreciation of fixed capital is included as an outlay in the accrual deficit, but does not show up in the cash deficit. A deficit calculated on the basis of national accounts (SNA) would be an accrual measure; the public sector borrowing requirement (PSBR) is measured on a cash basis. ^{11/ 12/}

In practice, countries’ deficit measures lie somewhere in between the complete cash and complete accrual measures. Even in countries which use a PSBR (cash) deficit concept, interest payments are usually measured as they accrue, rather than when actually paid. ^13/ On the other hand, revenues are almost always measured on a cash or quasi-cash basis because tax liabilities may be disputed and some percentage will never be collected. Finally, the accounting treatment of expenditures can make a significant difference to the measured deficit. The administrative procedures for executing government expenditures are complex, and take place in several stages from the time the government decides to undertake the outlay to the time the supplier considers himself paid. The size of the deficit can depend on the particular step at which expenditure is recorded as having taken place (IMF (1986), p. 87). ^14/ Comparison of deficits across countries with different recording practices, particularly during any fiscal year, can be misleading, as most expenditure commitments are made toward the beginning of the year, and most payments are made toward the end.

The economic analysis of accrual basis deficits is complicated by the institutional tendency to apply accrual accounting to the budget document rather than to the fiscal year. In other words, countries often prolong beyond 365 days the period over which transactions authorized in a given budget document may be carried out. (The extension is known as a “complementary period.”) Thus, in any fiscal year, transactions which change the measured deficit of the previous year can continue to take place alongside transactions determining the current year’s deficit. In such cases, the temporal common denominator for analyzing the budget balance together with other macro variables is lost.

A difficult problem in the measurement of the conventional deficit arises with attempts to reconcile the cash and accrual concepts. As noted by Diamond and Schiller (pp. 32, 42-44), if delays in payment are unanticipated, they represent forced borrowing from suppliers, with the result that the cash measure of the borrowing requirement misrepresents the sources of credit to government. If the delays are anticipated, suppliers will inflate their prices to compensate, and the government will pay a premium for its purchases.

The total reconciliation item between the cash and accrual deficits is often defined as arrears, but, because of justifiable lags in the expenditure process, because arrears may be run up through extrabudgetary expenditures, or because the emergence of arrears is often hidden by offsetting tax reliefs to suppliers, the definition of arrears is more complicated than a resolution of timing differences in expenditure recording. The existence of arrears that cannot be measured reduces the validity of the deficit as a measure of the government’s budget constraint or its impact on the economy.

Worse, countries with chronic liquidity crunches have developed formal procedures for turning arrears into longer-term debt instruments, which boost actual holdings of government debt above its long-run sustainable rate, inasmuch as they are usually held involuntarily by suppliers. The issue of chits or bonds in recognition of government’s debt to suppliers is recorded as a cash payment, and thus, inflates the recorded cash measure of the deficit compared with actual cash disbursements by government (see Table 3). ^15/

Table 3.

Burkina Faso: Consolidated Central Government ^1/

(In billions of CFA francs)

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/Burkina Faso is also a good example of a country where measured expenditures can be greatly affected by the existence of a complementary period. Expenditure from the 1985 budget during calendar year 1985 was CFAF 49 billion; total expenditure carried out in calendar year 1985 (including from past budgets) was CFAF 55 billion; and total expenditure from the 1985 budget (some of which took place in 1986) was CFAF 61 billion.

Table 3.

Burkina Faso: Consolidated Central Government ^1/

(In billions of CFA francs)

	1982	1983	1984
Reported cash balance less:	-6.2	0.5	-3.3
Deferred payment vouchers equals:	1.7	-1.3	1.1
True cash balance	-4.5	-0.8	-2.2

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/Burkina Faso is also a good example of a country where measured expenditures can be greatly affected by the existence of a complementary period. Expenditure from the 1985 budget during calendar year 1985 was CFAF 49 billion; total expenditure carried out in calendar year 1985 (including from past budgets) was CFAF 55 billion; and total expenditure from the 1985 budget (some of which took place in 1986) was CFAF 61 billion.

View Table

Table 3.

Burkina Faso: Consolidated Central Government ^1/

(In billions of CFA francs)

	1982	1983	1984
Reported cash balance less:	-6.2	0.5	-3.3
Deferred payment vouchers equals:	1.7	-1.3	1.1
True cash balance	-4.5	-0.8	-2.2

Source: IMF: Government Finance Statistics Yearbook, 1988.

^1/Burkina Faso is also a good example of a country where measured expenditures can be greatly affected by the existence of a complementary period. Expenditure from the 1985 budget during calendar year 1985 was CFAF 49 billion; total expenditure carried out in calendar year 1985 (including from past budgets) was CFAF 55 billion; and total expenditure from the 1985 budget (some of which took place in 1986) was CFAF 61 billion.

In the rest of this survey, it will be assumed that the conventional deficit is well-specified (following footnote 1 on p. 4), and the discussion of refinements to the conventional deficit will ignore the basic problem just discussed—that every budget speech refers to a different kind of deficit.

1. The current deficit: the impact of government on saving

The conventional deficit measures the difference between public investment and public saving. In order to isolate public (dis)saving, the current deficit assigns a weight of zero to investment outlays and capital revenues such as asset sales; that is, the current deficit is the difference between noncapital revenues and expenditures. The current deficit is of particular interest to economists because the lack of public sector capital budgeting appears to be shortcoming compared with accounting procedures of private firms: “if we maintained a separate and conceptually correct current and capital account system, the deficit on current account would be the true deficit, […because] for capital items, any excess of expenditures over receipts on capital account does not change the net asset position of the government, since the new debt is matched by a new government asset.” (Boskin (1982), p. 298.) Moreover, in the 1960s, it was commonly held that current expenditures should be fully financed by taxes, whereas, like a private firm, the government could legitimately finance its socially profitable investment by debt (Conklin and Sayeed (1983), p. 28). According to this view, the deficit on current account provided a measure of the extent the government strayed from “prudent management.”

Though the current deficit is intuitively simple, it has proved difficult to calculate for several reasons. First, the measure is useful primarily when comparing the government with the other components of the national accounts, or assessing government’s accounts according to the accounting norms of the private sector. However, detailed public sector accounts are usually first available on a financial basis, rather than on the accrual basis compatible with other sectoral and/or enterprise accounts. ^16/

Second, accounting concepts of investment are much narrower than the economist would like. For instance, most investment in human capital is considered a current outlay, despite its importance in explaining growth (Chelliah (1973), p. 749; Goode (1984), p. 240). Third, the current/capital mix of any “investment” project can be dissected (or, indeed, politically manipulated) in an almost infinite number of ways, to give many different measures of government saving.

Because of these operational problems, restriction of the deficit measure only to current account transactions has been infrequent. However, as externally financed programs of structural adjustment have become more prevalent, there has been renewed interest in the concept for developing countries. This is so because structural adjustment programs tend to disequilibrate the conventional balance, through temporarily large injections of subsidized lending for capital expenditure, and large one-time budget revenues from privatization. There is, therefore, pressure on the conventional deficit to widen, with the paradoxical implication that the structural adjustment has left the country even further from sustainable medium-term growth. To provide a more appropriate benchmark for judging these programs, it is argued that a deficit measure that excludes their temporary influences on the capital account will give a better indication of permanent adjustment efforts. The change in government saving, though clearly a rough proxy, has thus resurfaced as a summary gauge of the gains from structural adjustment (World Bank (1988)).

There are problems in using the current deficit as this kind of indicator. First, structural policies (such as tax reform) often involve J-curve effects (such as short-run revenue losses during the shift to the new tax system) which reduce government saving at the time of the adjustment program, though with the expectation of improving them in the medium term. Conversely, many structural reforms involve investments which imply heavy recurrent costs following completion, so that government saving may fall in the medium term. Finally, it may be difficult, if not illegitimate, to separate the disequilibria caused by a structural adjustment program from the “disequilibria” caused by other exogenous shocks or the business cycle. If so, a core deficit or a cyclically adjusted deficit (discussed in 5. below) might be a more precise measure of the extent of permanent adjustment.

2. The impact of government on aggregate demand

While the conventional deficit is an approximation of government savings minus investment, it is a poor measure of government’s contribution to aggregate demand. This inadequacy is evident from the Keynesian balanced budget multiplier—the simplest case of a weighted fiscal balance. More realistically, different elements of government expenditure and revenue generate different net increases to, and withdrawals from, demand. A measure of the budget’s overall impact on aggregate demand can be derived by weighting its components according to the expansion/contraction in aggregate purchasing power they represent—assuming these can be estimated.

Aggregate demand-based deficit measures have focused on the separation of exhaustive expenditures (on goods and services) and transfers. If private and public propensities to consume differ, it becomes important to identify the ultimate user of budgetary resources. Transfers such as pensions and unemployment benefits merely redistribute purchasing power from one part of the private sector to another. In terms of their impact on aggregate demand they are akin to negative taxes rather than to government’s expenditure on goods and services (Boskin (1982), pp. 296-97; Buiter (1985), p. 14; Bean and Buiter (1987), pp. 5-6). ^17/

Though transfers amount only to redistribution among different private sector groups, they will influence aggregate demand if the different groups have different marginal propensities to consume: that is, only in a simple case would the weight on transfers be the same as the weight on the taxes by which the transfers were financed. Borpujari and Ter-Minassian (1973) estimate the impact of the deficit on aggregate demand in an economy with three income groups: wage-earners, renters, and government transfer recipients.

The inclusion of transfers in government spending may overestimate government’s contribution to aggregate demand also because there are lags in how quickly transfers can be spent. This problem arises particularly often in economies with several layers of government: a transfer from central to local government may not increase aggregate demand until the year after it was recorded in the budget of the central administration. To address this problem, the elements of the annual budget may be weighted not only by their eventual impact on demand but also by the time it takes for that impact to become evident.

It should be noted that the use of a weighted deficit to analyze aggregate demand is peculiarly Keynesian. A more monetarist approach would argue that any impact of government on aggregate demand comes through the monetary financing of the deficit. If so, then, at least in a closed economy, all expenditures carry the same weight, and the liquidity balance (see 4. below) would be a more appropriate calculation. ^18/

3. The domestic deficit

If the economy is open, the full impact of government transactions may not be felt domestically. Since trade and capital flows between the public sector and the external sector vary enormously from country to country, a given conventional deficit can encompass a spectrum of contributions to domestic demand. For instance, expenditure on domestic goods that is fully financed by foreign grants increases aggregate demand with no offsetting withdrawal. Government imports financed by domestic taxes reduce aggregate demand by the full extent of the import bill—a case where government expenditure may have contractionary rather than expansionary effects on domestic demand. The overall deficit could well be zero in each of the two examples, though they each imply an opposite domestic impact. To isolate the effect of government on aggregate demand in an open economy, the “domestic deficit” must be separated from the “foreign deficit.”

The domestic deficit is measured by assigning nonzero weights only to those budgetary elements that directly affect the domestic economy. The foreign deficit—the impact of the budget on the balance of payments—can be measured by assigning nonzero weights to budget transactions only if they are directly connected to the external sector. (See, for instance, Borpujari and Ter-Minassian (1973), p. 815; Chelliah (1973), p. 770.) ^19/

Calculations of domestic and foreign deficits are particularly significant when the public sector itself has sizeable trade or capital flows to and from the rest of the world. Then, the overall deficit measure can be a particularly misleading indicator of the fiscal stance: for instance, devaluation may cause the budget deficit to widen if government imports or foreign debt service are large, suggesting an expansionary fiscal policy—though resources injected into the economy by government remain unchanged or may even fall.

Almost all applications of the domestic deficit measure have been carried out for oil exporting countries with a nationalized petroleum industry. Unless the monetary impact of oil receipts is sterilized, their use to finance expenditure will be expansionary but the conventional measure of the budget deficit would not predict the expansion. ^20/ Similar expansions can occur when foreign grants are large. Oil exporting country studies include Morgan (1979: 12 major oil exporters); Stillson (1979: Indonesia, Jordan, and Oman); Mackenzie (1981: Kuwait, Nigeria, and Saudi Arabia); Nasution (1989: Indonesia); Gil Díaz (1988: Bolivia); and Vaez-Zadeh (1989: Venezuela). Similar considerations apply to the case of foreign grants.

4. The liquidity balance

Morgan (1979), p. 82, shows that the domestic budget balance determines the direct effect of the government budget on money creation. However, strictly speaking, the relevant deficit measure, if government’s impact on aggregate demand comes from the way it finances its activity rather than from its tax and expenditure policies, is the liquidity budget balance—that share of government requirements that has to be financed by domestic credit. Resources financed from other sources would have a zero weight in the liquidity deficit. Hence, the liquidity balance differs from the domestic balance because it excludes domestic nonbank borrowing by government (which amounts merely to a rearrangement of private sector portfolios and not to money creation), while including central bank profits transferred to government and government interest payments to domestic banks.

Chelliah (1973), p. 774, notes a more restrictive definition of the liquidity balance: a measure of the impact of government on high-powered money. This definition would put zero weight on government credit from commercial banks.

5. Removing the effects of fluctuations in economic activity on the budget

While the budget deficit affects aggregate demand, aggregate demand also affects the budget deficit. Inter alia, income tax revenues will usually be lower and benefit transfers higher when unemployment is high. In other words, the budget deficit is affected by the business cycle, and the impact of discretionary policy changes may differ depending on at which stage of the business cycle they are implemented. Since the 1940s, but mainly in the 1970s, deficits abstracting from the impact of the business cycle have been calculated—that is, with weights of zero on the transitory/cyclical elements of revenue and expenditure. These measures have, in their heyday, been surveyed comprehensively (Blinder and Solow (1974), and Heller et al. (1986)).

There are two main classes of “permanent” or long-run deficits. The full-employment deficit (or structural balance) was derived in the belief that “a small surplus in that budget would ensure a high level of national saving while permitting built-in fiscal stabilizers to damp cyclical fluctuations” (de Leeuw and Holloway (1983), p. 27). ^21/ Notwithstanding its virtues, following a rule of full-employment budget balance could still imply the expansion of the public debt—since, on average, economies operate below full employment, so that, on average, expenditure would exceed revenue. Therefore, the cyclically adjusted budget balance was developed to provide a budget balance rule that would maintain a constant level of public liabilities. The methods of calculation of the two approaches have differed little:

“To construct a cyclically adjusted budget, the essential steps are (1) choosing a reference trend for GNP free from short-run fluctuations, (2) determining the responsiveness of each category of receipts and expenditures to short-run movements in GNP (e.g., cyclical tax elasticities), (3) applying these responses to gaps between trend GNP and actual GNP, and (4) adding the expenditures and receipts “gross-ups” from step 3 to the actual budget to obtain a cyclically adjusted budget. The first step, selecting a GNP reference trend, is the most important and controversial. Other things being equal, the higher the level of the reference trend, the smaller the cyclically adjusted deficit” (de Leeuw and Holloway (1985), p. 232).

The full employment deficit can then be defined as the cyclically adjusted balance when the GNP reference trend selected is potential output (Muller and Price (1984), p. 1). Potential output was defined as “the highest possible utilization of resources without accelerating inflation—an extraordinarily difficult concept to quantify” (de Leeuw and Holloway (1983), p. 27).

In this case, when neither taxes nor expenditures are a function of actual output, the budget balance measures only the effect of discretionary policy, that is, it carries to a more sophisticated level the distinction between policy decisions (above the line) and their endogenous implications (de Leeuw et al. (1980)). The change in the full-employment balance from year to year measures the fiscal impulse, that is, the effect, of fiscal policy (as opposed to all budget items) in total aggregate demand. ^22/

Attempts to measure the fiscal impulse have led to several intermediate estimates of budgetary stance that use potential output as a benchmark, for measuring revenue or (usually) expenditure (Dernburg (1975), p. 829; Heller et al. (1986); and Muller and Price (1984)). As a guideline for debt management, the cyclically adjusted or trend deficit is more appropriate than the full-employment deficit. In it, “neutral” expenditures and revenues are not estimated as functions of potential output but instead as functions of “average output.” In other words, estimation of the trend deficit separates those transactions that change long-run public liabilities from those that merely redistribute assets and liabilities within a cycle. While full-employment deficit estimates are cognizant of the impact of automatic stabilizers on deficit size, and adjust for them in various ways, the cyclically adjusted deficit includes the average value of the automatic stabilizer component of the deficit.

The trend deficit has been estimated by the Muller and Price (for the OECD), de Leeuw and Holloway, and Heller et al. (for the IMF). The OECD estimates the average automatic stabilizer directly, de Leeuw substitutes “mid-expansion” output for potential output, and the IMF removes the average automatic stabilizer implicitly through its subtraction of the fiscal impulse.

Trend deficit measures are subject to the disadvantage that they may not be equivalent to the “underlying” or “permanent” deficit in the economy, because they factor into the calculation of the trend transitory shocks. Particularly in countries undertaking structural reforms, these one-time disturbances could be equivalent to several percentage points of GDP. Tanzi (1982) has recognized this problem in his description of a variant of the trend deficit, the core deficit, which removes from the actual deficit not only cyclical influences but also one-time or transitory shocks (“such as temporary taxes, postponement of inevitable wage increases, building up of arrears and so on” (p. 6)). However, no published series of core deficit estimates exists.

It should be noted that a related branch of literature has developed, concerning the statistical techniques necessary to estimate the trend deficit. See, for instance, Cuddington and Urzua (1989). Trend stationary models are the most popular technique (Nelson and Plosser (1982)); however, because those are autoregressive moving average models, they would not succeed in estimating a core deficit, since they factor transitory shocks into the trend deficit.

Blanchard (1990) makes the point that deficit measures far simpler than trend deficits exist which distinguish between induced and discretionary fiscal policies, and which index sustainability. Trend deficits require needless assumptions about “whether there are cycles around a stable trend, (…) whether the economy will return to lower unemployment and so on” (p. 6). Instead, “[i]nduced changes in fiscal policy can be defined as those changes which come from changes in inflation, interest rates and output growth over the previous year—or over the previous ten year average—values. How to choose the benchmark is still a relevant question, but not one which requires taking a stand on where the economy will or should return.” Moreover, “[w]hat matters in terms of sustainability is where the country expects to be over the next three to ten years, not necessarily some mid-cycle point” (p. 7). Blanchard derives alternative easy-to-calculate measures, the simplest of which do not require forecasts.

6. The primary deficit: removing the effects of previous deficits on the budget

Although the structurally adjusted deficit is sometimes presented as measuring the impact of discretionary government policy, it includes an important nondiscretionary variable, namely, interest payments on the stock of public debt—which is usually predetermined by the size of previous deficits. The primary deficit (or “noninterest deficit”) attempts to measure the discretionary budgetary stance by attaching a zero weight to net interest payments in the budget (Barth et al. (1989)). ^23/ The primary deficit is also valuable as an indicator of the success of policies in moving the economy toward a sustainable growth path:

“The primary deficit measures how current actions improve or worsen the public sector’s net indebtedness, and it is important for evaluating the sustainability of government deficits. Although fiscal deficits can be run indefinitely, the primary balance must eventually become positive to cover at least part of the interest on current debt. If public revenue and the economy as a whole grow faster than the real interest rate, then even the primary balance can remain in deficit. However, it is generally not possible in the long run to always grow faster than the interest rate” (World Bank (1988), p. 56). ^24/

A comparison of the primary and conventional deficits in Table 4 illustrates the heavy burden of interest payments even in relatively stable economies, such as Spain, that have accumulated large public debts. Despite positive primary balances, conventional deficits remain.

Table 4.

Measures of the Fiscal Balance Under Alternative Treatments of Interest Payments ^1/

(In percentage of GDP)

Sources: Argentina, Brazil, Chile, and Mexico: Tanzi (1989); Israel: Bank of Israel; Ghana and Kenya: Thanos Catsambas and Miria Pigato (1989); and Spain: IMF staff estimates.

^1/Since data are obtained from different sources and country definitions may vary, the magnitudes are not comparable across countries.

Table 4.

Measures of the Fiscal Balance Under Alternative Treatments of Interest Payments ^1/

(In percentage of GDP)

	Conventional	Operational	Primary
Argentina
1983	-10.2	-10.2	-4.2
1985	-4.1	-4.1	-1.7
1987	-6.3	-5.6	-1.6
Brazil
1981	-13.0	-6.2	-4.8
1985	-27.9	-4.3	-0.6
1988	-45.3	-4.0	1.6
Chile
1983	-2.8	n.a.	1.0
1986	-1.9	-1.2	2.8
1988	3.6	3.8	8.0
Ghana
1981	-6.4	5.5	-4.3
1985	-2.7	-0.4	-1.2
1987	-0.3	-0.4	1.2
Israel
1985	-5.4	-1.6	10.0
1988	-4.9	-3.6	4.2
Kenya
1982	-6.5	-3.2	-2.9
1986	-5.4	-0.8	-0.5
1987	-7.6	-6.3	-2.9
Mexico
1981	-13.8	-10.8	-9.1
1985	-9.5	-1.0	3.3
1987	-15.9	2.0	5.0
Spain
1982	-5.6	…	-4.6
1985	-6.7	…	-3.2
1987	-3.6	…	0.1

Sources: Argentina, Brazil, Chile, and Mexico: Tanzi (1989); Israel: Bank of Israel; Ghana and Kenya: Thanos Catsambas and Miria Pigato (1989); and Spain: IMF staff estimates.

^1/Since data are obtained from different sources and country definitions may vary, the magnitudes are not comparable across countries.

View Table

Table 4.

Measures of the Fiscal Balance Under Alternative Treatments of Interest Payments ^1/

(In percentage of GDP)

	Conventional	Operational	Primary
Argentina
1983	-10.2	-10.2	-4.2
1985	-4.1	-4.1	-1.7
1987	-6.3	-5.6	-1.6
Brazil
1981	-13.0	-6.2	-4.8
1985	-27.9	-4.3	-0.6
1988	-45.3	-4.0	1.6
Chile
1983	-2.8	n.a.	1.0
1986	-1.9	-1.2	2.8
1988	3.6	3.8	8.0
Ghana
1981	-6.4	5.5	-4.3
1985	-2.7	-0.4	-1.2
1987	-0.3	-0.4	1.2
Israel
1985	-5.4	-1.6	10.0
1988	-4.9	-3.6	4.2
Kenya
1982	-6.5	-3.2	-2.9
1986	-5.4	-0.8	-0.5
1987	-7.6	-6.3	-2.9
Mexico
1981	-13.8	-10.8	-9.1
1985	-9.5	-1.0	3.3
1987	-15.9	2.0	5.0
Spain
1982	-5.6	…	-4.6
1985	-6.7	…	-3.2
1987	-3.6	…	0.1

Sources: Argentina, Brazil, Chile, and Mexico: Tanzi (1989); Israel: Bank of Israel; Ghana and Kenya: Thanos Catsambas and Miria Pigato (1989); and Spain: IMF staff estimates.

^1/Since data are obtained from different sources and country definitions may vary, the magnitudes are not comparable across countries.

7. The operational deficit: removing the effects of inflation from interest payments

The interest bill is beyond the control of current fiscal policy, not only because it represents the cost of previous deficits, but also because monetary policy can affect interest rates and hence budgetary interest payments. In addition, fluctuations in inflation can significantly change the size of government nominal debt service.

Inflation affects the budget in many ways. First, inflation can alter the real flow of budgetary resources. For example, inflation can increase real tax liabilities in an unindexed progressive tax system by shifting taxpayers into higher tax brackets (“fiscal drag”). On the other hand, by lengthening collection lags, an acceleration of inflation might reduce real tax revenues (see Tanzi (1987)). Such inflation-induced real changes are a legitimate source of budgetary disturbance, and should not be removed from any meaningful deficit measure. ^25/

Second, inflation may have “balance sheet” effects, that is, it can change government net worth by altering the real value of its assets and liabilities, even though revenue and expenditure flows remain unaffected. Unrealized valuation effects are not captured by the annual flow measures discussed in this section, and are dealt with in Section V.

Finally, inflation can lead to confusion in interpretation when the real and nominal elements of a transaction represent different components of the budget, but are not disentangled by conventional budgetary accounting practices. In particular, inflation reduces the real value of the outstanding stock of unindexed public debt but may compensate creditors for such erosion in their real assets through higher nominal interest rates. In other words, some of the government’s interest payments on its debt are in reality part of the amortization of that debt. If the inflationary component of interest rates is not removed from the interest bill, the deficit will be overstated by the size of the amortization element included as interest payments above the line, rather than below.

The magnitude of the deficit overstatement depends on the size of domestic debt outstanding and on its terms and denomination. When public debt is held in the form of short-term floating-interest domestic bonds, the growth rate of nominal interest payments will in general exceed the inflation rate, and, as inflation rises, the conventional deficit will expand in terms of GDP, even if inflation has no other impact on the budget. It is simple to show that, with floating-interest debt, the deficit/GDP ratio is a positive function of inflation and of the initial debt/GDP ratio. ^26/

A consequence of this relationship is that countries with identical inflation, debt/GDP ratios, and ratios of tax revenues and noninterest expenditures to GDP may, nevertheless, show very different conventional fiscal deficits, depending solely on the composition of their debt. These shortcomings of the conventional deficit under inflation, and the validity of alternative measures to alleviate the problem have been analyzed, among others, by Catsambas (1988), Cukierman and Mortensen (1983), Eisner (1984), Eisner and Pieper (1984), Gil-Díaz (1986), Miller (1982), and Tanzi, Blejer, and Teijeiro (1988).

The most popular alternative suggested to alleviate the problem is the operational deficit, which places a zero weight on the inflation-induced portion of interest payments. It is defined as the conventional deficit minus that part of the debt service which compensates debt holders for actual inflation, that is, the primary deficit plus the real component of interest payments. Real interest payments are the interest bill less the inflation-induced erosion in the value of the debt (measured, for example, by the outstanding debt times actual inflation). ^27/ In some countries which had high inflation, such as Brazil and Mexico (Table 4), immense differences arise between conventional and operational deficits. Moreover, trends in the two alternative measures can diverge markedly. In both Brazil and Mexico, the conventional deficit indicates rapidly growing imbalances during the 1980s, whereas the operational deficit signals some improvement in Brazil and a remarkable adjustment in Mexico.

In order to compare the merits of the operational deficit with the conventional measure, it is necessary to review its economic rationale. The operational deficit excludes inflation-induced interest payments on the assumption that they are similar to amortization payments in their effects on the economy—namely, that they do not represent new income to recipients, and are willingly reinvested in government bonds, at existing market conditions, and therefore they do not affect the level of aggregate demand in real terms. Real interest payments, on the other hand, can be consumed without reducing a bondholder’s net wealth, and thus have a similar expansionary impact as any other type of expenditure. The relative usefulness of the two deficit measures reduces, thus, to the question of how inflation-induced interest payments are spent: are they used to buy new bonds or to finance consumption? In other words, does rising inflation erode the real demand for government bonds, or is the sustainability of the public debt invariant to inflation? More precisely, full rollover of inflation-induced interest service as assumed by the operational deficit requires, first, a stable real demand for bonds, and, second, that the rate of inflation is not an argument in the demand function for government bonds, which is affected only by real interest rates.

The irrelevance of inflation from real demand for bonds, though essentially an empirical question, is supported by portfolio theory. ^28/ There are, however, some channels through which rising inflation could reduce the real demand for bonds—for instance, a higher nominal stock of debt and a high rate of inflation could raise the perception of default risk. ^29/

If inflation were to reduce the real demand for bonds, then, in an economy with accelerating inflation, inflation-induced interest payments would not be fully refinanceable under existing market conditions but would require either higher real interest rates or higher bond liquidity, thus increasing demand pressures. The operational deficit measure excluding the inflation component of interest payments would then underestimate the degree of fiscal imbalance.

There are also technical problems in the calculation of the operational deficit. For instance, the choice of the price index is not straightforward and there are presentational difficulties when interest rates are negative in real terms, in which case the conventional deficit measure would have to be adjusted downward by a magnitude greater than actual interest payments. Furthermore, the operational deficit has a macroeconomic deficiency: by correcting the deficit for the impact of inflation on it, we lose the ability to assess the impact of the deficit on inflation. Despite these difficulties, the operational deficit provides useful information to policymakers when the inflation rate is very high. In principle, it is a lower-bound estimate for the public sector deficit, relevant when full rollover of broadly defined amortization is realistic.

1. The traditional scope of government

It is tempting to think of traditional government as a pyramid, from the apex of the central decision maker(s), through the central administration, down through the numerous regional, municipal and local governments. However, the different parts of traditional government are distinguished through more dimensions than are representable in a pyramid. ^31/ Governments, even at the central level, include a plethora of public agencies which cannot easily be ranked “above” or “below” one another, such as investment boards, industrial development authorities, utility regulating bodies, space research laboratories, social security funds, etc. Only rarely are their powers neatly subordinated to an overseeing ministry.

The difficulty of constructing a generally applicable organizational structure for government is such that it may be impossible to derive an exhaustive list of entities that should be taken into account in arriving at an undisputable figure for the fiscal deficit. Nonetheless, there are some guidelines—completeness of coverage and of consolidation, and the recognition of nonfinancial governmental activities—that are important for determining the scope of government.

2. Public production and trade

A strict definition of government presupposes a restricted array of public sector economic activities, merely the provision of nonmarket goods and services and the redistribution of income. However, the concepts of public sector and government increasingly diverge the more involved in production and commerce publicly owned entities become. While the activities of public enterprises, marketing boards, and other publicly owned entities that produce or trade are to some extent motivated by profit, they have special characteristics, in many cases monopolistic, with prices and, sometimes, quantities being primarily the result of government policy. Clearly, the effects of these activities should be factored into a meaningful measure of the public sector balance. However, this poses a number of conceptual and methodological problems including the appropriate measurement of the fiscal component of public enterprises’ activities.

There are two aspects specific to the assessment of the fiscal impact of public enterprises which have a bearing on the proper measurement of budgetary balance. ^33/ First the enterprises that comprise the public sector must be selected, and, second, that portion of their operations which has a fiscal impact must be identified.

The choice of entities for inclusion in the public sector depends on the distinctions between government and enterprises and between “public” and “private.” In IMF (1986), the criterion used to distinguish between general government and nonfinancial public enterprises is not legal or institutional, but rather the nature of the activities they perform. The distinction comes both from the nature of the goods and services they supply and the differing character of their revenues: taxes are compulsory levies while income from market sales is essentially voluntary. ^34/ In general, when the unregulated market cannot be expected to generate the optimal provision of the good or service but some price mechanism could still be set in motion, public enterprises rather than government may be called upon to execute a desired intervention.

The operational distinction between “public” and “private,” which determines the extent to which transactions by entities outside general government should be incorporated in measures of fiscal activity rather than classified as private sector operations, is more difficult to pin down. ^35/ A common criterion is the simple ownership principle: any enterprise is considered “public,” if direct and indirect government participation in its equity exceeds 50 percent. This legal concept is not really satisfactory because it does not assess the degree of governments actual control over the enterprises’ decisions, nor does it evaluate the weight of public policy objectives in entities’ operations. A different approach is taken by Stella (1989) who looks at the overall impact of enterprises on the public finances and on net worth transfers. The operations of many enterprises, privately as well as publicly owned, are supported by a variety of state guarantees, tax benefits, or other types of financial assistance. If these interplays of government financing and enterprise operations are such as to raise private sector net worth, these results are akin to government deficits in their effect on private perception of wealth and hence on consumption. In this sense, their operations could therefore be considered part of the public sector.

A unifying, although not very operational, criterion for classification comes from the “soft budget constraint” concept developed by Kornai. It defines “public” firms as being immune from bankruptcy and therefore unconcerned with covering costs. ^36/ Soft budget constraints, with their contingent claim on budgetary resources, are the proximate reason for linking some of the operations of public enterprises with those of the government for an accurate assessment of global fiscal impact.

In addition to the explicit macroeconomic impact of enterprise policies which further government objectives, government backing gives the suppliers, creditors, and customers of enterprises security that their client will remain in business or that, if it fails, its debts will be paid. This security could have an impact on resource allocation through a crowding-out effect, by shifting resources away from other firms toward industries with government backing.

One approach for analyzing the macroeconomic dimension of public firms is to evaluate the budgetary impact of their operations--through the flows between public enterprises and the government (Premchand (1983)) and through the effects of public enterprises on the overall volume of government revenue, expenditure, and public investment (e.g., Ramanadham (1984), Chapter 3). A second approach, perhaps more valid, is to consolidate the relevant part of public enterprises’ operations with the rest of the government budget. Consolidation, however, raises a number of important methodological measurement issues (Stella (1989)). The central question is how to define and measure enterprise revenue and expenditure in a way that is compatible with government’s concepts. Clearly, enterprises’ gross sales revenue is not comparable to tax revenue, nor should the purchase of inputs be added to current budgetary expenditure. It could be argued, however, that public enterprise prices contain implicit subsidies and taxes that will be reflected in profits and losses and that these profits and losses are the financial flows that should be consolidated since they closely correspond to budgetary inflows or outlays. ^37/

A more difficult problem arises when profits are not explicitly transferred to the Treasury or they are lower than potential competitive profits owing to hidden subsidies. If the enterprise is making losses, the subsidy element may remain obscured through the enterprise’s ability to borrow from domestic and foreign sources other than the government.

Practically, consolidation also faces the problem that the structure of government and enterprise budgets are very different:

“The government budget is usually subdivided into receipts, expenditure, and borrowing. The budgets of enterprises are mostly organized, like commercial budgets, on a dual basis, viz., revenue and capital” (Premchand (1983)).

In addition, the statistical base of the accounts differs. ^38/ Almost always, enterprise accounts are on an accrual basis while cash accounts are the usual budgetary standard. Stella (1989) claims that this difference cannot be resolved simply by converting enterprise flows into cash accounts because, in assessing the impact of the public sector on the economy, it is indeed more correct to measure enterprise activities, like any other business activities, on an accrual basis since this gives a truer reflection of performance. When capital expenditure is important, the divergences between cash and accrual accounting can be significant. ^39/

In sum, the issue of how to measure the gross flow of government-like activities of public entities remains unresolved. In the case of a marketing board, for example, one would not want to amalgamate the gross value of purchases with treasury outlays, nor would one want to combine the gross value of sales receipts with tax revenue. The policy element is only the subsidy or tax implicit, if any, and the quantitative measure of the subsidy is the difference between buying and selling prices as reflected in the operating position of the enterprise.

3. The quasi-fiscal operations of central banks

In many countries, the distinction between the responsibilities of the Treasury and the central bank has become blurred, with the latter performing “quasi-fiscal” activities—not specifically connected with monetary policy. These activities are diverse: they include the management of explicit subsidies, debt service, and transfers, the provision of preferential credit, the bailout of ailing industries, etc. It has frequently been argued that these quasi-fiscal operations are similar to other budgetary activities and should be included in a comprehensive measure of the public sector balance. Particularly important, analytically, is the central bank’s implicit levy of taxes, either through the exchange rate system, ^40/ or through the imposition of unremunerated reserve requirements.

There are many difficulties in separating the central bank’s monetary from its quasi-fiscal activities. ^41/ Moreover, differences in accounting practices (e.g., cash versus accrual) raise consolidation problems akin to those with nonfinancial public enterprises. In a methodological study of central banks’ fiscal activities, Robinson and Stella (1988) start from a benchmark case: they claim that central banks that have operating profits and transfer them fully to the Treasury ^42/ do not distort the conventionally measured deficit even if they perform quasi-fiscal activities, provided that these activities only affect the central bank’s profit-and-loss accounts during the budget year in question. Deviations from this benchmark would require an adjustment.

The two most important departures arise: a. from quasi-fiscal activities that change the composition of the central bank’s balance sheet (rather than the profit-and-loss account); and b. when the central bank makes a loss which is covered by an equivalent reduction in its net worth. ^43/

a. A prominent quasi-fiscal activity which entails a change in central bank’s balance sheet composition is its lending to the private sector for public policy purposes. An important example is preferential sectoral lending, financed by high-powered money. Because these loans could be very similar to budgetary loans, there is an argument for their inclusion as a deficit-determining item analogous to government’s net lending. However, the gross incorporation of all central bank lending to the private sector into the fiscal deficit would clearly be misleading. Much of central bank lending is done for pure monetary—and not fiscal—reasons. This is the case with rediscounting, open-market operations, and sterilization. It is quite evident that such intervention for monetary purposes should not result in an increase in a consolidated fiscal deficit measure. To preserve distinctions among types of central bank lending, the ideal solution would be to transfer quasi-fiscal lending from the central bank to government’s accounts, with a counterbalancing change in net credit to government from the central bank. ^44/

Central banks’ balance sheets can also be affected by capital gains and losses arising from valuation changes. This may take place when the central bank is forced to take over private (or public enterprise) debt or to rescue troubled financial institutions. Another common source of valuation changes is the change in the value of the central bank’s net foreign-exchange holdings, which could arise from external parity fluctuations or from a domestic devaluation that changes the domestic currency counterpart of net foreign assets, resulting in an accounting capital profit or loss. ^45/

There is no clear view on how these valuation changes should be treated in relation to the fiscal stance. ^46/ Robinson and Stella distinguish between unrealized and realized gains. They claim that unrealized gains should be excluded from central bank profits because they attract no new resources—that is, they are not revenue-enhancing—while the expenditure “financed” by them is a deficit-determining item similar to other expenditure financed by central bank credit. Should the unrealized gains become realized, Robinson and Stella claim that:

“compared with the situation that would have obtained with no revaluation gain, purchasing power in the private economy is reduced by the amount of the valuation gain, and thus expenditure “financed” by realized gains is similar to expenditure financed from revenue. If the central bank’s accountants took note of the capital gain…transfers to the government would increase, reducing the fiscal deficit.”

b. The second deviation from the initial benchmark arises when central banks make losses. Significant central bank deficits are frequent in developing countries, sometimes exceeding conventional fiscal deficits (Table 5). Reasons for these losses vary. Their most common causes are quasi-fiscal, such as the requirement on central banks to lend without interest or at very low interest rates for policy purposes. Operational losses also arise from the administration of a multiple exchange rate system (which may include an implicit subsidy to preferred buyers) and from currency devaluations when the central bank has net foreign exchange liabilities vis-à-vis the domestic sector. ^47/

Table 5.

Central Bank Deficits ^1/

(In percentage of GDP)

Sources: Argentina, Julio Piekarz (1987); Costa Rica, Ana Rodríguez Aguilera (1987); Ghana and Kenya, Catsambas and Pigato (1989); Philippines, IMF staff estimates; and Uruguay, Dionisio Onandi and Luis Viana (1987).

^1/Since the indicators are taken from different sources, they are based on various definitions of the concept of central bank losses and thus are not strictly comparable.

Table 5.

Central Bank Deficits ^1/

(In percentage of GDP)

	1984	1986
Argentina	-2.5	-1.6
Costa Rica	-4.3	-3.8
Ghana	-2.1	-0.6
Kenya	-3.8	-5.4
Philippines	-5.2	-2.8
Uruguay	-4.2	-4.0

Sources: Argentina, Julio Piekarz (1987); Costa Rica, Ana Rodríguez Aguilera (1987); Ghana and Kenya, Catsambas and Pigato (1989); Philippines, IMF staff estimates; and Uruguay, Dionisio Onandi and Luis Viana (1987).

^1/Since the indicators are taken from different sources, they are based on various definitions of the concept of central bank losses and thus are not strictly comparable.

View Table

Table 5.

Central Bank Deficits ^1/

(In percentage of GDP)

	1984	1986
Argentina	-2.5	-1.6
Costa Rica	-4.3	-3.8
Ghana	-2.1	-0.6
Kenya	-3.8	-5.4
Philippines	-5.2	-2.8
Uruguay	-4.2	-4.0

Sources: Argentina, Julio Piekarz (1987); Costa Rica, Ana Rodríguez Aguilera (1987); Ghana and Kenya, Catsambas and Pigato (1989); Philippines, IMF staff estimates; and Uruguay, Dionisio Onandi and Luis Viana (1987).

^1/Since the indicators are taken from different sources, they are based on various definitions of the concept of central bank losses and thus are not strictly comparable.

Whether or not central bank losses arise from quasi-fiscal activities, there is a case for their explicit inclusion in the public sector deficit. In addition to their expansionary effect, there is an issue of symmetry of treatment. Although it is common practice to transfer central bank profits to government, thus reducing the fiscal deficit, current losses do not elicit a transfer from the government to the central bank, so the measured deficit does not rise. To prevent measurement biases, the impact of central bank losses should be included in the public sector balance by recording, for example, a budgetary transfer or a subsidy from the government, thus properly increasing the recorded fiscal deficit.

To summarize, ideally, government accounts should incorporate quasi-fiscal revenues and expenditures, leaving central bank accounts covering only monetary activities. A second-best solution would be, first, that central bank operational losses be consolidated into the fiscal deficit by the addition of a transfer from government to the central bank financed by credit from the central bank. Second, an estimate of the size of central bank quasi-fiscal activities falling outside the profit-and-loss account should be made, and then amalgamated into the adjusted fiscal deficit. Such a hybrid deficit would mix net worth with cash concepts, but would have value as a supplementary indicator showing the approximate impact of central bank quasi-fiscal activities on the overall public sector balance.

4. The budgetary dimension of the public financial sector

Typically, public financial institutions are excluded from the coverage of the public sector and are consolidated with the private banking system. However, these institutions often engage in a multitude of activities (such as preferential credit allocations, subsidized interest rates, etc.) with a clear fiscal content. To the extent that such activities go beyond pure liquidity management which could have been carried out by private financial intermediaries, it is possible that the exclusion of resources provided by the public financial system from the measured public sector balance creates a misleading impression of the fiscal policy stance.

As in the case of nonfinancial enterprises, in order to differentiate between public and private financial institutions, it would be appropriate to consider the implications of their operations on the distribution of income and wealth. Using this approach, pure commercial banks in which the government owns a large, or even a majority stake, should not be considered “public” if their activities have nothing to do with public policy and if they fully finance their operations at prevailing market conditions. The relevant public sector would, therefore, include only those public institutions such as development banks, sectoral credit institutions, mortgage banks, building and loan associations, finance and investment companies, as well as insurance companies and pension funds, which mobilize savings through the receipt of contractual premia but invest in assets frequently selected by public policy considerations. ^48/

Why should some or all of the activities of these institutions be considered quasi-fiscal? The answer appears to be related to the nature of their operations on both sides of the capital market. Public financial institutions are, in many countries, perhaps the most common means of directing credit for policy purposes. In several countries, they account for more than three quarters of all long-term finance. (See World Bank (1989).) Therefore, the same considerations that apply to direct budgetary net lending by the government and to quasi-fiscal lending by the central bank seem to apply here.

There is, however, an important difference. Unlike the government and the monetary authority, but like typical private financial institutions, many public ones act as intermediaries financing at least part of their long-term financial claims by selling long-term financial assets to the public. Nevertheless, public institutions clearly operate in the capital market under special conditions. They were created to provide services that, for whatever reason, other institutions had found not worthwhile or too risky to provide. Thus, public institutions are likely to be less profitable, and be more exposed to risk than other financial institutions, and at a disadvantage in mobilizing voluntary resources from the financial markets. Their survival, hence, often depends on government guarantees (giving them an edge in the market) or on explicit government subsidies, monopoly power over market segments, preferential access to government-mobilized resources, or other forms of preference or protection, including exclusive access to external loans. ^49/ This being the case, the operations of these institutions would exert crowding-out pressures on financial markets not conceptually different from those arising from the financing of other government activities, and therefore should not be neglected when assessing the overall economic impact of the consolidated public sector. ^50/

1. Intertemporal shortcomings of the conventional deficit

Developments in private sector consumer theory have been paralleled (albeit with a lag) by changes in our understanding of public sector behavior. It was always clear that the public sector (being less liquidity-constrained than any private individual) did not finance its expenditure completely out of current income. However, several recent developments in the world economy have highlighted the fact that the government, even if infinitely lived, is constrained—like private consumers--by the size of its permanent income.

The debt crisis has shown that there are perceived limits on governments’ ability to repay borrowing from future generations to finance present consumption and the U.S. social security debate has generated awareness of the implications for today of government commitments to spend or repay tomorrow. The conclusion that governments face an intertemporal budget constraint not unlike that of private agents cannot be avoided. It has also become clear that governments’ consumption path is determined by wealth as well as by income: privatization programs that seemed to improve the financial position of public sectors have shown that governments can dissave to finance consumption in any period. Finally, it is now recognized that governments’ consumption paths can be importantly affected by price and valuation changes. This has been amply illustrated by the effect on governments’ financial position of swings in the value of the dollar over the 1980s, the various Latin-American hyperinflations and the development of debt buyback schemes through which governments have profited by the fall in value of their debt. ^51/

Some deficiencies in traditional measures of the deficit become particularly evident when government behavior is recast in an intertemporal rather than annual framework—and when attention is shifted from short-run demand management to the sustainability of the deficit. Deficiencies include the omission of valuation adjustments, the treatment of asset sales, and of the financial implications of entitlement programs and government guarantees. Specifically, the problems are as follows:

a. The conventional deficit includes no provision for valuation changes in government assets or liabilities, though these could conceivably change the sign of the budget balance in any fiscal year. One facet of this issue has already been discussed: adjustments to the deficit that separate amortization from interest payments on public debt in inflationary regimes are a partial recognition of the impact that prices can have on the nominal deficit. However, government’s ability to pay can also be affected, in real terms, by inflation, devaluation, changes in the terms of trade or in relative prices, and real capital gains or losses on the purchasing power implicit in government assets and liabilities, though none of these effects are captured by a summary of government transactions during that fiscal period. ^52/

b. Conventional deficit measures usually include receipts from privatization and the sale of other assets as a revenue item. Structural programs or pressures to cut the flow deficit have resulted in the conversion into liquid assets of nonfinancial tangible and even intangible assets that were not previously considered. When assets such as land, embassies, or aircraft are sold, they provide immediate cash to alleviate the current year’s financing burden. The amounts can be important--and help to overcome drastic temporary downturns in tax revenue (as in the case of Argentina during its recent hyperinflation (Table 6)). However, the government is worse off by the replacement cost of the assets (arguably their realized market sale value; Goldsmith (1985), p. 92).

Table 6.

Argentina: The Sale of the Tokyo Embassy ^1/

(In percentage of GDP)

Source: Centro de Estudios Macroeconomicos de Argentina, Buenos Aires (unpublished).

^1/GDP ratios for 1988 are quarterly averages for the full year; 1989 ratios are for the second and fourth quarters.

^2/Includes proceeds from asset sales.

Table 6.

Argentina: The Sale of the Tokyo Embassy ^1/

(In percentage of GDP)

	1988	1989 (II)	1989 (IV)
Current revenue	17.6	13.0	18.2
Capital revenue 2/	0.9	3.8	0.3

Source: Centro de Estudios Macroeconomicos de Argentina, Buenos Aires (unpublished).

^1/GDP ratios for 1988 are quarterly averages for the full year; 1989 ratios are for the second and fourth quarters.

^2/Includes proceeds from asset sales.

View Table

Table 6.

Argentina: The Sale of the Tokyo Embassy ^1/

(In percentage of GDP)

	1988	1989 (II)	1989 (IV)
Current revenue	17.6	13.0	18.2
Capital revenue 2/	0.9	3.8	0.3

Source: Centro de Estudios Macroeconomicos de Argentina, Buenos Aires (unpublished).

^1/GDP ratios for 1988 are quarterly averages for the full year; 1989 ratios are for the second and fourth quarters.

^2/Includes proceeds from asset sales.

The nature of the problem asset sales pose for the deficit differs depending on whether the assets disposed of have previously been purchased by government through the budget or whether they have “always” formed part of the public patrimony (for instance, in the case of mineral rights). Treating as revenue, the sales of previously purchased investment goods in computing the measured deficit is justified by the unorthodox treatment of capital expenditure in government accounts. Unlike private sector capital (and the treatment of public capital in the SNA) which is depreciated over its lifetime, public capital is fully expensed in the fiscal year it is purchased. ^53/ This merging of the current and investment accounts, which makes consistent the inclusion of the full value of an asset sale as a revenue item, can be justified when looking at the annual financing needs of government (Stella (1989), pp. 19-22), but is, however, a misleading indicator of the sustainability of the government’s policy stance.

The inclusion of revenues from assets other than investment goods as an “improvement” in government’s ability to pay is incorrect by any private sector accounting practice. When the government sells land or mineral rights, for example, it has merely changed the composition of its portfolio: it has the cash but it no longer has the asset. If it earned the market value of the asset, then it is no better or worse off than prior to the sale. ^{54/ 55/}

c. The conventional deficit can be severely affected by “revenues” which create liabilities for the future or “expenditures” which represent the liquidation of past liabilities. On the revenue side, the traditional deficit often includes changes in the net position of social insurance programs. However, social insurance contributions supposedly confer entitlements on contributors and as such commit the government to higher future spending. Thus, social security contributions do not represent free-and-clear revenues, and their inclusion in the deficit overstates government’s ability to pay. On the other hand, because they are contingent claims (contingent not only on contributors’ attaining old age, or ill health, but also on changes in government legislation), the magnitude of outlays they will eventually require is difficult to determine. ^56/

Analogously, the conventional deficit can be dramatically inflated in any year by government’s payment of previously guaranteed debt, or insurance contracts, such as exchange guarantees or bail-outs of underwritten entities (like insolvent public enterprises, or the U.S. savings and loan industry). In reality, such payments are stock adjustments—the sum of the accumulated risk costs borne by government over the life of the guarantee. Unlike the private sector, which mitigates the impact of bad debts by accumulating loan loss reserves as offsetting stocks, the government usually fails to make provision for expected defaults. Hence, the costs of risk-bearing are not spread out over the life of the risk, but are charged only upon realization of the risk’s downside.

The measurement problem in the conventional deficit is not just that meeting current entitlements or paying up for past guarantees boosts the deficit, but also that, at any time, the conventional deficit provides an over-optimistic indicator of government’s long-run ability to pay, because it does not factor in the expected future cost of entitlements and contingent liabilities assumed by government. Moreover, the calculation of the expected cost of contingent claims is complicated by the possibility of moral hazard: even if the entitlements and guarantees are not funded or provisioned against, the assumption of liability by government may change private sector behavior. Eisner (1990) clearly rephrases the problem:

“It may be pointed out that loan guarantees or deposit insurance indirectly finance real spending just as they might if treasury expenditures were made up front. In a sense, the explicit and implicit deposit insurance or guarantees raised the budget deficit at the time the S&Ls made the loans that ultimately turned bad…the expenditures were made then. They then financed the now half-empty office buildings or homes worth only a fraction of their construction costs. Current government borrowing to finance the purchases of S&L assets only makes explicit an element of deficit or debt that was implicit earlier in the commitment of backing to S&L liabilities.”

Towe (1989), p. 2, takes these problems one step further, recasting them in terms of their implications for budgetary control:

“…since the issuance of such contingencies may not impact the current budget, while having severe cash-flow implications for the future, there may not exist sufficient controls, under conventional accounting constraints, to maintain the level of such contingent liabilities at a prudent level.”

Clearly, appropriate accounting for contingent claims requires an intertemporal framework.

2. The deficit as an indicator of government solvency: changes in public sector net worth

The so-called deficiencies described above have one thing in common: they do not affect the current year’s borrowing requirement. ^57/

Moreover, while the consequences of these issues generate ample debate, their combined effect on aggregate demand in any single year would be well-nigh impossible to measure. Hence, it should be stressed that the main reason for tackling these difficulties is in order to refocus the deficit measure as an indicator of the long-run sustainability of government policy—put dramatically: of the solvency of government.

According to Bean and Buiter (1987), p. 27:

“A government is solvent if its spending programme, its tax-transfer programme and its planned future use of seigniorage are consistent with its outstanding, initial financial and real assets and liabilities (in the sense that the present value of its spending programme is equal to its comprehensive net worth).”

In other words, over its lifetime, while a government can shift consumption between periods by saving and borrowing, it will be unable to consume more than its total income plus its initial endowment. ^58/ Under this definition, the “fiscal deficit” would be equivalent to the dissaving of government in any year (reduction in its net worth).

Like the net worth of a firm, the net worth of government is specified in its balance sheet, and the overall fiscal deficit in any period is equal to the difference in balance sheets at the beginning and end of the period. The methodological and measurement difficulties which bedevil the specification of the government balance sheet—far more than the firm’s—are discussed below.

b. An ideal government balance sheet

Buiter (1983, especially p. 310); also Bean and Buiter (1987, p. 28ff); describes the ideal “comprehensive consolidated public sector balance sheet at current market or implicit prices.” To capture the complete array of ways in which government can increase or run down its net worth in a global balance sheet, government assets should include: financial assets; real capital—including nonmarketable social overhead capital, equity (mainly in public enterprises—partly marketable); land and mineral assets (discovered and undiscovered—partly marketable); the present value of the future tax program (including social security contributions); and the imputed present value of seigniorage. Liabilities would include government debt (domestic and foreign, indexed or not); the stock of high-powered money; and the present value of social insurance and other entitlement programs (including guarantees). Government net worth is then the balancing item.

While Buiter’s construct provides a clear conceptual framework defining government net worth, it is far from operational. Even at the conceptual level, the definitions of capitalized values of tax and spending programs are subject to enormous controversy. And the valuation of tangible assets presents special difficulties when it must be undertaken on the massive scale necessary to encompass complete public sector holdings. Moreover, since public assets are less frequently traded than private assets, their prices may be difficult to identify. Indeed, were public assets traded, their prices and those of their currently traded substitutes might be very different from private sector prices in a thinner market (not augmented by government purchases and sales). ^60/

Despite these problems, valuable work has been done on many items in the comprehensive balance sheet. In particular, Eisner and Pieper (1984), Eisner (1986), and Boskin, Robinson and Huber (1987a), present improved balance sheets containing many innovations which address the deficiencies in deficit measurement detailed above. ^61/ Specifically, as discussed below, efforts have been made: (1) to assess the magnitude of valuation changes in financial net assets, for a more accurate picture of government liquidity; (2, a) to provide a more economically correct estimate of capital formation and the capital stock, by applying a more realistic depreciation scheme than the current system of annual expensing; (2, b and c) to provide a more comprehensive picture of government’s ability to pay by including in the balance sheet public land and mineral rights; and (3) to create a framework for assessing the eventual impact of contingent claims on the budget. However, the remaining element of the comprehensive balance sheet, the present value of the tax program (4), presents conceptual difficulties large enough to cast doubt on the interpretation of any measure of government net worth.

(1) The valuation of financial assets

Budget deficits have been considered damaging, in an intertemporal sense, because they add to the public debt and thereby erode the sustainability of the government’s expenditure path at current levels of tax revenue. However, as Eisner (1984), p. 140 points out:

“The “underlying reality…that every dollar of deficit…adds a dollar to debt” is simply not true in a real sense if prices are not constant. And if interest rates fluctuate, the statement is not true even with reference to the market value of nominal debt.” In particular, positive inflation rates erode the real value of public debt, so that governments which are net debtors can have rising net worth while continuing to run deficits. Moreover, increasing interest rates erode the market value of previously issued fixed-interest debt.

Thus, to arrive at the change in net worth attributable to changes in the values of (net) financial assets, the change in their nominal par value from one balance sheet to the next should be augmented by two adjustments--the difference between the real and nominal values of net financial holdings, and the difference between their face value and their market value at the time the net worth calculation is being made.

These adjustments have been more widely applied than any other balance sheet reconciliation item, because, even when economists were not concerned directly with net worth measures, they were troubled by the discrepancy between measures of net government spending and measures of changes in net government liabilities (see Muller and Price (1984), p. 8; Eisner and Pieper (1984), p. 12). Adjusted series for public debt appear in Miller (1982); Eisner and Pieper (1984; recalculated in Boskin et al, 1987a); de Leeuw and Holloway (1985); and Eisner (1986). Eisner (1976) presents revaluation estimates for a range of government assets and liabilities.

Russo (1987), p. 12, however, has objected to the par-to-market adjustment, on the grounds that (save if the government were to raise taxes in order to prepay its debt) the public debt is always amortized at its face value; neither gains nor losses from shifts in market valuation over the life of the loans are ever realized. ^62/ Hence, he claims, such shifts, however large their effect may be in any year, are irrelevant to the consideration of the sustainability of the deficit.

(2) The valuation of real assets

While some valuation problems are common to all assets, specific issues arise in the valuation of depreciable assets, land and mineral rights.

(a) Real capital and depreciation

Because the capital stock is estimated by accumulating annual government capital formation, ^63/ it is sensitive to the form of depreciation assumed across vintages of capital, that is, to the assumption of the rate of net investment by government. The impact of different depreciation schemes on estimates of the capital stock is discussed in Boskin, Robinson, and Roberts (1985a). Of course, the validity of any depreciation scheme depends on how closely it approximates economic depreciation. Boskin et al. (1985b) apply a geometric depreciation scheme with rates inferred where possible from the ratio of new to used asset prices, on the argument that “Equipment depreciates faster than straightline in the early years, and structures depreciate more slowly.” (p. 16) ^64/ Goldsmith, Ott, and Austin, Eisner, and the BEA use straightline depreciation in their calculations, while Kendrick (1976) uses double-declining depreciation.

(b) Land

Methods of land valuation have been of concern to policymakers since governments started to collect taxes, and a wide literature exists at the microeconomic level. The problem for the government balance sheet is one of aggregation: the information required for the micro-oriented techniques is too detailed to be applied to all public sector holdings. There are also pitfalls in making global inferences from partial data: for many reasons, public sector land (such as military land) may not be a close substitute for private land; and, as mentioned above, were all public land marketable, land prices might be very different from what they are at present.

In fact, as Boskin et al. (1985b), pp. 931-32 point out, global estimates of the value of U.S. federal government land (in Goldsmith (1985), Milgram (1973), Ott and Yoo (1980), Eisner and Pieper (1984), and Boskin et al. are each simply extrapolations (using different combinations of price indices, and adjustments for changes in total acreage and in land composition) of a 1946 estimate made by J.E. Reeve et al. (1950; quoted in Boskin et al.). “These studies […] demonstrate how successive refinements of basic data often hang by a very slender thread.” (p. 931) “A new benchmark estimate for the value of federal land in a particular year is especially important.” (p. 935).

As to the valuation of government land in other countries, Goldsmith (see. p. 119) cites difficulties in valuing nonagricultural land, which is often consolidated with the value of the buildings erected on it (so that a proportional valuation factor must be assumed), and in assessing the share of forest on so-called agricultural land.

(c) Mineral rights

The inclusion of mineral rights in the government balance sheet is arguably even more important than the inclusion of government land, because changes in the pace of their direct exploitation, sale or lease are seen by governments as an important way of improving their short-term financial position, and are therefore a prime generator of the problem mentioned earlier, whereby the sale of an asset/exhaustible resource gives a misleadingly optimistic picture of government wealth accumulation, by not offsetting the revenue by the cost of the depletion of the asset. ^65/ It is also true, however, that the large fluctuations in oil prices observed over the last two decades could create much volatility in government net worth from year to year, if applied directly to valuing the stock of mineral rights--unhelpful volatility since only a small portion of stocks would be sold.

One way or other, as Boskin et al. (1985b), p. 924 point out, no work was done prior to their pioneering study on the valuation of federal mineral rights; and we have seen no applicable study in other countries. Boskin et al. estimate expected unproven as well as proven gas and oil reserves—in the spirit of Buiter’s forward-looking comprehensive public sector balance sheet. ^66/ The inclusion of undiscovered reserves is important for the correct interpretation of government revenues, because the lease of mineral rights typically begins with the sale of exploration rights to unproven fields. The government earns revenue (bonuses) by exploiting firms’ expectations about reserves, even if the fields prove to be dry; and, as before, the revenues are not free and clear but come from the government’s having ceded an (expected) asset.

Two measurement complications make accounting for exhaustible resources more difficult than accounting for the government’s capital stock. First, stocks of undiscovered reserves must be recalculated each time discoveries are made—and relationship between proven and unproven reserves may not be linear. Second, the inclusion of an estimate for mineral rights with estimates of the value of land is problematic, because it is not clear to what extent the value of land already internalizes the value of the minerals underneath. ^67/ Ignoring these complications, Boskin’s work (Table 7) gives an idea of the implications of changes in the value of real assets for the fiscal deficit. ^68/

Table 7.

United States: Influences on Federal Net Worth

(In billions of current U.S. dollars)

Sources: Line 1: United States (1989a, Table B-79); Lines 2 and 3: Boskin et. al. (1985b).

^1/Federal Government receipts less expenditures on a National Income and Product Accounts basis.

^2/Line 4 is illustrative only; it has not been checked for inconsistencies in definition.

Table 7.

United States: Influences on Federal Net Worth

(In billions of current U.S. dollars)

		1979	1980
1.	NIPA balance 1/	-16.1	-61.3
2.	Change in value of federal land	+17.2	+36.9
3.	Change in value of oil and gas rights	+93.8	+208.8
4.	Augmented balance (lines 1 + 2 + 3) 2/	+94.9	+184.4

Sources: Line 1: United States (1989a, Table B-79); Lines 2 and 3: Boskin et. al. (1985b).

^1/Federal Government receipts less expenditures on a National Income and Product Accounts basis.

^2/Line 4 is illustrative only; it has not been checked for inconsistencies in definition.

View Table

Table 7.

United States: Influences on Federal Net Worth

(In billions of current U.S. dollars)

		1979	1980
1.	NIPA balance 1/	-16.1	-61.3
2.	Change in value of federal land	+17.2	+36.9
3.	Change in value of oil and gas rights	+93.8	+208.8
4.	Augmented balance (lines 1 + 2 + 3) 2/	+94.9	+184.4

Sources: Line 1: United States (1989a, Table B-79); Lines 2 and 3: Boskin et. al. (1985b).

^1/Federal Government receipts less expenditures on a National Income and Product Accounts basis.

^2/Line 4 is illustrative only; it has not been checked for inconsistencies in definition.

(3) The valuation of entitlements, contingent claims and guarantees

Particularly in the United States, the proper treatment of social security obligations in the fiscal accounts has generated much discussion (see, for instance, Rosenbaum (1990)). Towe (1989), p. 10 ff. describes the main options, from the most restrictive method (the accumulated benefit-cost approach) to that most comparable to net worth (the actuarial balance). While these accounting treatments have been developed mainly for social security programs, their application can be considered for much broader ranges of entitlement schemes and insurance programs.

The accumulated benefit-cost approach to valuing the net impact of an entitlement/insurance program is used in the private sector, where the expected liability of the program is defined only with respect to current participants, and according to current rules (see also Boskin et al. (1987b), p. 44). This approach would narrowly restrict the consideration of contingencies (and therefore of government solvency) to the question of whether present participants will continue to pay their expected subscriptions/premia and become eligible (for example, by living long enough) to collect their expected benefits.

The somewhat less restrictive actuarial fairness approach to valuation defines the deficit or surplus in a contingency program as the difference between the (aggregated) expected present value of the payouts to each of a program’s participants over the program’s duration and the expected net present value of their payments, thus allowing consideration of expected changes in policy and participation. “Fairness” requires that over the lifetime of each participant, the program must be in balance.

Actuarial balance requires that expected (present value) payments to all present and future participants be equivalent to total expected contributions (adjusted for operating expenses and any relevant endowment or reserve). If the former exceeds the latter, the program has a negative net worth. Boskin et al. (1987b) present estimates of the U.S. social security balance based on this criterion calculated over 75 years. However, they use these estimates to illustrate the extreme sensitivity of such present value calculations to assumptions about contingencies: “[M]oving all of the economic and demographic projections from intermediate to either optimistic or pessimistic [assumptions] results in a change which is larger than the privately held national debt.” (p. 45)

The calculation of program deficits under any of the above criteria also requires an estimate of probabilities. Degrees of certainty in payments can vary widely between programs, and have been used as classification criteria--distinguishing between, for instance, pension schemes, where expected outcomes are smooth and predictable once the demographics have been identified, and deposit insurance to financial institutions, where the risks are highly correlated, leading with a small probability to extremely high payouts. (Boskin, et al., (1987b) especially p. 15.) Moreover, risks may be even higher than guarantees or premia paid would suggest, if political or other pressures force government to treat non-insured agents on a par with insured agents during a systemic crisis. ^69/ Boskin et al. (1987b) derive backward-looking estimates of probabilities for defaults on loans from the Small Business Administration, but caution that “In the case of an insurance program […], where the risks of default across borrowers are highly correlated and very rare, a model based on historical experience can be misleading.” (p. 32)

The approaches described measure only the first-order present value of the contingency program. Thus, according to these criteria, all programs in which guarantees are issued without charge (often the case with exchange guarantees (Robinson and Stella (1988), p. 29), are deemed to be in deficit—though the government would not have issued them without the expectation of some social benefit (such as risk spreading). The value of the social benefit might conceivably be estimated (in some cases, by comparing costs in a market without the guarantees (Wattleworth (1988), p. 58), and imputed to the government accounts, but it will usually be impossible to assess the impact of the social benefit on other elements of the government balance sheet. ^70/

A final point made by Towe concerns the treatment of reserves sometimes set up to finance contingency programs. While these reserves would seem to represent an offset to any calculation of a deficit in the program, they will do so only when not held in the form of other government liabilities.

(4) The valuation of the present value of the tax program

Eisner (1984), pp. 139-40 takes the view that changes in the value of contingent claims are likely to be met by changes in taxes (or other redistributory legislation, and hence, that the inclusion of such claims in the deficit could give a misleading measure of the fiscal stance, out of line with the private sector’s perception of its claims on government. But if Ricardian equivalence is broadly defined, this view could be generalized to all potential reductions in government net worth, and the present value of the tax program simply replaces net worth as the balancing item in the comprehensive balance sheet.

Even if Ricardian equivalence does not hold, the government’s power to control its long-run net worth through altering tax and expenditure legislation ^71/ suggests that, even if government has a negative net worth given today’s policy package, it is not insolvent in the private sector sense, but merely must adjust the tax program by the amount of its “permanent deficit” ^72/ in order to return to sustainability. The indeterminacy of the net worth measure inherent in the flexibility of the government’s power to tax is the main philosophical problem with balance sheet or net worth concepts of the deficit. Given this indeterminacy, it is not clear that net worth measures can be constrained to be any less arbitrary than are flow measures.

In the limit, thus, government’s control over resources encompasses all of private sector income and wealth as well. Obviously, the sustainability of government policy would then depend on its impact on the total wealth of the economy--in other words, on private agents’ view of their net worth. ^73/

Kotlikoff (1989) recognizes this broadest of interrelationships in his proposal to substitute a “Fiscal Balance Rule” for present indicators of budget sustainability:

“[The Fiscal Balance Rule] says take in net present value from each new young generation an amount equal to the flow of government consumption less interest on the difference between (a) the value of the economy’s capital stock and (b) the present value difference between the future consumption and labor earnings of existing older generations. …[O]ne can use existing data to check whether it is being obeyed and, therefore, whether future generations are likely to be treated better or worse than current generations.” (p. 2).

In other words, if the present labor force pays for government consumption by taxes augmented by its interest earnings on the capital stock net of that part which finances dissaving by the old, government policy will not run down the economy’s capital stock and future generations will be as wealthy as past generations. Under this criterion, the fiscal deficit would be defined as government consumption in excess of taxes plus interest.