Appendix I Major Fiscal Policy Rules

This appendix surveys existing and proposed constraints on fiscal policy that can be defined as fiscal policy rules (see Section I).⁸⁵

Budget Balance

Probably the best known fiscal policy rules are those involving balance between government revenue and expenditure. This can be specified as the overall balance, the current balance, or the operating balance to be met each fiscal year. Alternatively, it can be defined over a longer period, in terms of a structural balance or a cyclically adjusted balance.

In the postwar period, in several industrial countries, governments were subject to various balanced-budget rules. In Germany, Japan, and the Netherlands, the rule was defined in terms of the current balance—government limited to borrowing capital expenditures only—commonly called the “golden rule.”⁸⁶ Subsequently, in the Netherlands, the government was committed to a structural deficit ceiling. Among developing countries, Brazil‘s balanced-budget requirement applies only to the central government, although certain forms of borrowing are treated as revenue.

According to the Maastricht Treaty, members of the European Union wishing to participate in Stage 3 of EMU (monetary unification, effective 1999) are required to contain their general government deficit at a level not in excess of 3 percent of GDP by 1997, following a convergence plan under way since 1992. In addition, the Stability and Growth Pact calls for “a medium-term budgetary position close to balance or surplus,” subject to the 3 percentage point reference value for the deficit in any year, so as to “allow for automatic stabilizers to work, where appropriate, over the whole business cycle.”⁸⁷ Largely accrual-based recording standards have been issued for this purpose, under the so-called excessive deficit procedure,⁸⁸ and compliance with this requirement will be verified ex post, after the end of the calendar year.

In New Zealand, the authorities are required to ensure that once a prudent public debt-GDP ratio (see below) is reached, it is maintained, on average, over a reasonable period of time, through balance between operating expenditure and revenue of the public sector.⁸⁹ The requirement allows for short-term cyclical deviations—albeit without specifying whether due only to automatic stabilizers or to discretionary actions as well—from the balanced-budget position. Furthermore, tax rates are required to remain stable over time, implying that the adjustment should take place on the expenditure side.

In Switzerland, the authorities have proposed constitutional amendments whereby the federal government finances would be balanced over the business cycle. The balanced-budget amendment would become effective in 2001, at which time the authorities would have achieved balance.⁹⁰ It is understood that the lower levels of government would cooperate in such an endeavor in order to halt the increase in the public debt-GDP ratio. In the United States, the balanced-budget amendment to the constitution, proposed on several occasions (in 1982, 1995, 1997) but so far rejected, would require the government to balance the federal budget each fiscal year. The rule could be waived only by a three-fifths majority of each house of Congress or in the case of an armed conflict or a threat to national security. The rule thus precludes both an explicit role for automatic stabilizers and concrete guidelines on how the rule is to be met.

A similar constitutional amendment is under consideration in Costa Rica. Intended for implementation by 1999, this amendment would limit the overall deficit to 1 percent of GDP or less. Coverage of the rule would extend to the entire public sector, including the public financial institutions, to capture sizable quasi-fiscal activities. Although details of implementation remain to be worked out, it is clear that a consensus has emerged to end the politically induced fluctuations in fiscal performance that have characterized Costa Rican public finances for more than two decades.

Among countries with federal systems, sub-national levels of government in Germany and in the United States are subject to the so-called golden rule: current revenue and current spending must balance each fiscal year. Nonbank borrowing is permitted to finance investment projects under certain well-defined conditions. In Germany, most Lander are subject, in principle, to the golden rule. In practice, however, there has been considerable latitude in defining current and capital expenditures. Moreover, the rule is applied to the approval of the budget rather than to the execution. In the United States, all state governments (with the exception of Vermont) follow the current balanced-budget rule under varying degrees of stringency. Whereas in certain U.S. states it is sufficient to enact a balanced budget, in others it is also necessary to implement it. A number of states do not permit carryover of unspent appropriations or of payables from one fiscal year to the next to meet the rule. Some states prescribe the creation and utilization of contingency reserves, and most states impose various kinds of limits on the amounts and types of debt that may be issued.

More recently, also at the subnational level, in Canada, six provincial governments (Alberta, Manitoba, New Brunswick, Nova Scotia, Quebec, and Saskatchewan) and two territories (Northwest and Yukon) have enacted balanced-budget rules. They range in stringency from requiring actual overall balance or limiting deficits (equivalent to 1 percent of expenditures or revenues) each year to prescribing current balance over a four- or five-year period. Escape clauses, permitting deficits, can be invoked in the event of an emergency or disaster (or in one case, a significant revenue fall beyond the control of the authorities). Some governments are required to observe the rule only on an ex ante basis; for others, the obligation extends to realized budget outcomes as well.

Borrowing

Some of the oldest functioning fiscal rules consist of prohibition of or limits on government borrowing. The borrowing constraint usually specifies the source of financing (central bank or all domestic sources) and the level of government (national or subnational) to which it applies.

Most of the advanced economies and some developing economies prohibit direct central bank financing of the general government as well as the rest of the nonfinancial public sector. Under the Maastricht Treaty, such a rule went into effect at the beginning of Stage 2 for all EU member countries participating in EMU. Normally, this rule leaves the extension of short-term advances to the government to the discretion of the central bank, as evidence of the bank‘s independence. The rule is somewhat less commonly found in developing countries and economies in transition.⁹¹ Under a strict variant, in Chile and Ecuador, both direct and indirect access to central bank credit is prohibited. Instead of outright prohibition, in some developing and transition economies (in the CFA franc zone, Brazil, Egypt, Morocco, the Philippines, and the Slovak Republic), central bank credit is limited to a proportion (usually between 5 and 20 percent) of government revenue in the preceding year.⁹²

Perhaps a better known rule prohibiting domestic government borrowing operates in Indonesia, having been introduced in the mid-1960s in the wake of an external payments crisis and high inflation. The coverage excludes certain off-budget operations, particularly those financed with oil export receipts or borrowing from abroad.

In many countries, there are limits on borrowing by lower levels of government.⁹³ In a number of them, borrowing is at most permitted for investment (Germany, Switzerland, and the United States) or as a proportion of revenue (Canada) or expenditure (Korea). In some developing countries (The Bahamas and Chile), borrowing by subnational governments is not permitted; in others (Mexico), the restriction applies only to external borrowing. Recently, in some economies in transition (Albania, Bulgaria, Poland, and Romania), borrowing prohibitions have been imposed in the face of weak overall institutional controls on local government finances.

Debt

A fiscal policy rule may consist of a limit on, or a target for, the stock of public debt as a proportion of GDP. At present, under a broad definition, the following two cases can be considered rules.

In the European Union, the Maastricht Treaty calls for a ceiling on the general government gross debt, set at a reference value of 60 percent of GDP, for participation in EMU. However, there is room for interpretation, especially in determining whether a country is making sufficient progress toward the reference value. The ceiling permits using privatization revenue for gross debt reduction—even though such an operation leaves net worth, with properly valued assets, unchanged. Furthermore, excluded from the debt ceiling are publicly guaranteed liabilities contracted by the rest of the public sector or the private sector, accrued liabilities under government pension schemes, and other contingent liabilities, including unfunded social security obligations for future pensions and health care benefits.

In New Zealand, the authorities are required to announce a medium-term plan for reducing the public debt ratio to a prudent level.⁹⁴ Accordingly, the previous government declared its commitment to reduce the net public debt to 20 percent of GDP in the medium term, as well as to maintain an adequate level of public sector net worth so as to provide a buffer against future unanticipated adverse developments. Coverage of the net worth component of the target extends to all outstanding liabilities, including the net actuarial value of government employee pensions under defined-benefit schemes and other quantifiable contingent liabilities and commitments, though excluding unfunded liabilities associated with public pension programs and other nonquantifiable liabilities. In Canada, as part of balanced-budget legislation, three provinces have imposed debt-reduction requirements. These provisions are specified in terms of an explicit target (Alberta) or are to be determined by the government within broad guidelines (New Brunswick and Quebec).

Reserves

In some ways analogous to a debt rule, a fiscal rule may prescribe a target accumulation of reserves (presumably in terms of liquid assets) for a future contingency. The contingency may take the form of old-age protection, insurance for depletion of a natural resource, or insurance against an unanticipated fall in key commodity prices. Such a rule is usually limited to a target contingency reserve ratio, in the context of an extrabudgetary fund.

To provide minimum funding for future pension claims in the United States, the social security trust funds (Old-Age, Survivors‘ and Disability Insurance, or OASDI) are targeted to accumulate contingency reserves equivalent to 100–150 percent of annual benefit payments. According to recent projections, this threshold reserve-benefit ratio is forecast to be maintained only until about 2025, unless changes in benefits or payroll contribution rates are made. In Canada, the Canadian Pension Plan is required to reach and maintain an annual reserve-benefit ratio of at least 200 percent. Under a new financing plan, the ratio is projected to increase to 500 percent over the long term. In Germany, the public pension reserve must maintain at least one month of benefits.

Instead of setting a minimum or target level of reserves, Chile‘s Copper Compensation Fund is required to accumulate excess revenue from copper exports, as a function of the difference between a medium-term reference price (determined by a six-year moving average of the export price) and the current export price of copper, and of various other factors. In the event of negative price differentials, transfers are usually made from the fund‘s reserves to the budget.

Implicit Rules

Implicit fiscal policy rules can be broadly derived from monetary or exchange rate rules. Perhaps the most common rule is a fixed nominal exchange rate—and, to a lesser extent, a preannounced crawling peg—which requires that the budget be kept close to balance, particularly given a fully open capital account and an undeveloped government securities market. Many IMF member countries—under the Bretton Woods system, the entire membership—have been, or currently are, relying on such an exchange rate rule, which imposes a high degree of fiscal discipline. However, the experience of Mexico, which until end-1994 held on to a preannounced crawling peg, illustrates that even a nearly balanced budget may not be sufficient to ensure the sustain-ability of a strict exchange rate rule when it becomes overvalued.

Much less frequent is a currency board arrangement,⁹⁵ which permits, in principle, government borrowing from the domestic banking system in addition to external and domestic nonbank sources but only at a high cost in terms of the crowding out of private activity, as the monetary expansion must be backed by foreign exchange inflows. Accordingly, in Argentina and Estonia, under recently established currency board arrangements, public finances (inclusive of privatization receipts) are to be kept close to balance, given limited access to alternative (domestic nonbank and external) sources of finance.

Other Rules

Specialized fiscal rules—not the main focus of this paper—applied to certain categories of government revenue or expenditure have their own rationale. A limitation, for example, on primary expenditure at constant prices or on the share of primary expenditure to GDP is usually driven by an effort to contain or reduce the size of government in the economy.⁹⁶ Rules requiring the earmarking of revenue for particular purposes or mandating a certain composition of expenditures over time (such as limiting defense outlays) are usually predicated on microeconomic grounds and may be difficult to maintain over an extended period.

Although rules on the composition of tax revenue or of primary expenditure may be arbitrary or inconsistent with rules encompassing aggregates, their adoption is justified in certain circumstances. Rules requiring balance between complementary outlays, for instance, on primary and secondary education—or between current (operations and maintenance) and capital expenditures on health care facilities or on highways—have a clear economic justification. The earmarking of certain user charges for the provision of specific services (for example, toll roads) or of payroll contributions for specific social insurance funds can be useful in creating both support and accountability for such programs. Rules on minimum tax rates—for example, value-added tax rates within the EU—within a federal or confederate context may be necessary to prevent undesirable tax competition among jurisdictions. In addition, explicit revenue-sharing arrangements—including apportionment rules for the corporate income tax base among subnational jurisdictions—or formula-based equalization transfers among different levels of government are intended as a distributive instrument among different tax jurisdictions.

Appendix II Macroeconomic Effects of Fiscal Restraint

An important rationale for fiscal policy rules is the elimination of the deficit bias experienced in many countries over the last two decades or so. This appendix summarizes the theoretical context and empirical findings as regards the effects of fiscal retrenchment on interest rates and growth, which are of particular relevance in ascertaining the likely consequences of fiscal rules.

At an elementary level, a hypothetical increase in government saving, in the form of reduced government purchases or lower disposable income (from higher taxes or reduced transfers), has a negative direct effect on aggregate demand. An ex ante increase in government saving leads to an increase in national saving, lower interest rates, a crowding in of private investment, and higher output in the long run.⁹⁷ As fiscal rules are expected to mitigate or remove the deficit bias inherent in the fiscal policy of the industrial countries, it is useful to examine existing empirical estimates of the macroeconomic consequences of fiscal retrenchment.

There are several factors that determine the quantity and quality of these effects, as reflected in fiscal multipliers, including monetary and exchange rate policy, openness of the economy, the form of fiscal adjustment (that is, government consumption, transfers, and taxes), and exogenous credibility effects that alter the risk of default. Standard Keynesian models with backward-looking expectations, developed in the 1970s, yield negative short-run effects, with multipliers often greater than 2 that last several years.⁹⁸ In more recently developed models that incorporate intertemporal budget constraints and rational expectations, fiscal multipliers reflect various (often opposite) effects (Table 3). Although the negative direct demand effect is still present, the total effect critically depends on the expectation of future policy action. If fiscal adjustment is perceived as permanent, expected increases in future output, as well as expected declines in future interest rates and tax rates, tend to encourage present investment and consumption, while mitigating the withdrawal of demand. For the same fiscal shock, the multiplier can exhibit substantial variation (between negative and positive values), depending on expectations regarding the future course of policy. In general, multipliers tend to be smaller if the fiscal policy adjustment is gradual but credible. The positive credibility effects of a continuous fiscal reduction (lower long-term interest rates and higher future income growth) are anticipated and thus realized today. These points have been illustrated by two recent fiscal simulations for Germany and the United States in the context of the IMF‘s MULTIMOD model.⁹⁹

Table 3.

Summary of Fiscal Multipliers

¹These models include DRI (Data Resources International), Compact Model of the European Economic Commission, the World Econometric Model of the Japanese Economic Planning Agency (EPA), the Project LINK World Model, the Multi-country Model (MCM) of the Federal Reserve Board, the INTERLINK Model of the Organization for Economic Cooperation and Development, and Wharton Econometrics Forecasting Associates World Model.

²These models include the McKibbin-Sachs Global Model (MSG), a two-country simulation model developed in the Research Department of the IMF MINIMOD), the Multi-country Model of John Taylor, and Patrick Minford’s Liverpool Model.

³These models include Global Economic Model (GEM) of the National Institute for Economic and Social Research, INTERMOD of the Canadian Department of Finance, the Multi-country Model (MCM) of the Federal Reserve Board, and the U.S. Economy Model (MPS) of the Federal Reserve Board.

⁴The models include INTERMOD of the Canadian Department of Finance, MULTIMOD of the Research Department of the IMF, the Multi-country Model of John Taylor, McKibbin-Sachs Global Model (MSG), and a three-country model developed in the Division of International Finance of the Federal Reserve Board (MX3).

⁵Patrick Minford’s Liverpool Model.

Table 3.

Summary of Fiscal Multipliers

(Normalized to I percent of GDP)
		Expenditure Increase		Tax Decline
Author	Simulation	Short term	Long term	Short term	Long term
IMF (1996b)	5 percent debt increase in the United States	1.1	-0.6	0.7	0.2
IMF (1995b)	10 percent debt increase in the United States	…	…	0.7	-0.2
	10 percent debt increase in Italy (increase in interest premium)	…	…	0.4	-0.2
IMF (1996d)	Gradual and credible deficit increase in Germany for total of 1 percent after four years (expenditure increase, partly offset by tax increase)	-0.3	-0.7	…	…
	Same but not credible	0.1	-0.7	…	…
	Front-loaded expenditure increase	0.5	-0.7	…	…
Bryant and others (1988)	Adaptive-expectations models1	1.2 to 2.0	0.0 to 1.9	…	…
	Rational-expectations models2	0.7 to 1.6	0.2 to 0.7	… …
Bryant, Hooper, and Mann (1993)	Adaptive-expectations models3	1.0 to 1.3	-0.13 to -0.01	…	…
	Rational-expectations models4	0.65 to 1.5	-0.09 to 0	…	…
	Monetarist rational-expectations model5	0.10	0.01	…	…

¹These models include DRI (Data Resources International), Compact Model of the European Economic Commission, the World Econometric Model of the Japanese Economic Planning Agency (EPA), the Project LINK World Model, the Multi-country Model (MCM) of the Federal Reserve Board, the INTERLINK Model of the Organization for Economic Cooperation and Development, and Wharton Econometrics Forecasting Associates World Model.

²These models include the McKibbin-Sachs Global Model (MSG), a two-country simulation model developed in the Research Department of the IMF MINIMOD), the Multi-country Model of John Taylor, and Patrick Minford’s Liverpool Model.

³These models include Global Economic Model (GEM) of the National Institute for Economic and Social Research, INTERMOD of the Canadian Department of Finance, the Multi-country Model (MCM) of the Federal Reserve Board, and the U.S. Economy Model (MPS) of the Federal Reserve Board.

⁴The models include INTERMOD of the Canadian Department of Finance, MULTIMOD of the Research Department of the IMF, the Multi-country Model of John Taylor, McKibbin-Sachs Global Model (MSG), and a three-country model developed in the Division of International Finance of the Federal Reserve Board (MX3).

⁵Patrick Minford’s Liverpool Model.

View Table

Table 3.

Summary of Fiscal Multipliers

(Normalized to I percent of GDP)
		Expenditure Increase		Tax Decline
Author	Simulation	Short term	Long term	Short term	Long term
IMF (1996b)	5 percent debt increase in the United States	1.1	-0.6	0.7	0.2
IMF (1995b)	10 percent debt increase in the United States	…	…	0.7	-0.2
	10 percent debt increase in Italy (increase in interest premium)	…	…	0.4	-0.2
IMF (1996d)	Gradual and credible deficit increase in Germany for total of 1 percent after four years (expenditure increase, partly offset by tax increase)	-0.3	-0.7	…	…
	Same but not credible	0.1	-0.7	…	…
	Front-loaded expenditure increase	0.5	-0.7	…	…
Bryant and others (1988)	Adaptive-expectations models1	1.2 to 2.0	0.0 to 1.9	…	…
	Rational-expectations models2	0.7 to 1.6	0.2 to 0.7	… …
Bryant, Hooper, and Mann (1993)	Adaptive-expectations models3	1.0 to 1.3	-0.13 to -0.01	…	…
	Rational-expectations models4	0.65 to 1.5	-0.09 to 0	…	…
	Monetarist rational-expectations model5	0.10	0.01	…	…

¹These models include DRI (Data Resources International), Compact Model of the European Economic Commission, the World Econometric Model of the Japanese Economic Planning Agency (EPA), the Project LINK World Model, the Multi-country Model (MCM) of the Federal Reserve Board, the INTERLINK Model of the Organization for Economic Cooperation and Development, and Wharton Econometrics Forecasting Associates World Model.

²These models include the McKibbin-Sachs Global Model (MSG), a two-country simulation model developed in the Research Department of the IMF MINIMOD), the Multi-country Model of John Taylor, and Patrick Minford’s Liverpool Model.

³These models include Global Economic Model (GEM) of the National Institute for Economic and Social Research, INTERMOD of the Canadian Department of Finance, the Multi-country Model (MCM) of the Federal Reserve Board, and the U.S. Economy Model (MPS) of the Federal Reserve Board.

⁴The models include INTERMOD of the Canadian Department of Finance, MULTIMOD of the Research Department of the IMF, the Multi-country Model of John Taylor, McKibbin-Sachs Global Model (MSG), and a three-country model developed in the Division of International Finance of the Federal Reserve Board (MX3).

⁵Patrick Minford’s Liverpool Model.

In addition, the magnitude of a fiscal retrenchment may have an influence on its macroeconomic effects; a large fiscal adjustment is more likely to signal a change in regime and modify consumers‘ expectations than a small one. Through a descriptive examination of large-scale fiscal adjustment experiences over the last 25 years in OECD countries, as well as through more formal econometric tests, it has been shown that such episodes frequently result in smaller output losses than suggested by standard Keynesian models and, in a number of cases, in negative fiscal multipliers.¹⁰⁰ This non-Keynesian effect of large-scale fiscal retrenchment on activity cannot be systematically ascribed to monetary policy or to cross-country spillover effects, but rather it appears to reflect consumer response to the size of fiscal policy shocks.

Concerning the nexus between fiscal adjustment and interest rates, empirical findings from both reduced-form estimations and structural model simulations (Table 4) corroborate that interest rates are sensitive to a change in the public debt ratio. Specifically, an increase in government debt, equivalent to 25 percent of GDP, would result in an increase of 125 to 500 basis points in long-run interest rates.¹⁰¹ These should be regarded as a lower bound of the beneficial long-run effects that may result from debt reduction arising from the introduction of a permanent rule limiting debt or deficit ratios.¹⁰² The understatement of the interest rate effect stems from the inability of underlying models to capture the credibility effect associated with the operation of a binding rule, in addition to the interest rate decline attributable simply to a onetime increase in government saving. The credibility effect of a fiscal rule on interest rates can perhaps be approximated by a reduction in the default risk premium in financial markets on the debt of a country that pursued a vigorous fiscal adjustment program to correct an unsustainable path (Denmark and Ireland in 1980s, Italy in the early 1990s). Alternatively, in an attempt to measure default risk, it was found that interest differentials on comparable government bonds among EU member countries, normalized to a common currency, rarely exceeded 50 basis points, displaying a statistically insignificant relationship between interest rates and debt ratios.¹⁰³

Table 4.

Summary of the Effect of Changes in Public Debt Ratio on Long-Term Interest Rates

Table 4.

Summary of the Effect of Changes in Public Debt Ratio on Long-Term Interest Rates

Author	Type of Estimation or Simulation	Change in Interest Rate Due to 25 Percent Increase in Public Debt-GDP Ratio (basis points)
Tanzi and Fanizza (1995)	Panel estimation	175
Ford and Laxton (1995)	Cross-section systems estimation (world debt as explanatory variable)	250–450
Helbling and Wescott (1995)	Estimation using aggregate world data	400–500
IMF (1996a)	MULTIMOD simulations	300
Faruqee, Laxton, and Symansky (1996)	Simulation of a structural closed-economy model	125

View Table

Table 4.

Summary of the Effect of Changes in Public Debt Ratio on Long-Term Interest Rates

Author	Type of Estimation or Simulation	Change in Interest Rate Due to 25 Percent Increase in Public Debt-GDP Ratio (basis points)
Tanzi and Fanizza (1995)	Panel estimation	175
Ford and Laxton (1995)	Cross-section systems estimation (world debt as explanatory variable)	250–450
Helbling and Wescott (1995)	Estimation using aggregate world data	400–500
IMF (1996a)	MULTIMOD simulations	300
Faruqee, Laxton, and Symansky (1996)	Simulation of a structural closed-economy model	125

Again, however, since none of these countries was following a balanced-budget or debt-reduction rule, there appears to be a tenuous relationship between fiscal control (measured by debt levels) and default risk. Of course, the large domestic interest differentials may be influenced by the degree of fiscal control, but they are more likely to reflect inflation or exchange rate risk differentials. In yet another attempt to measure the exogenous credibility effect, similar fiscal adjustment scenarios were simulated for Italy and the United States with one important difference: the default risk premium for Italy was assumed to decline as the adjustment took place, whereas it remained unchanged for the United States. The results show that compared with the United States, fiscal adjustment in Italy would tend to appreciate the domestic currency, interest rates would decline significantly more, and output would be more buoyant.¹⁰⁴

Appendix III Stochastic Simulations of Fiscal Rules

A central question that arises in connection with fiscal policy rules relates to their impact on the short-term variability of aggregate output and income. To examine this question, a number of stochastic simulations, applying the IMF‘s MULTIMOD model, have been performed for the Group of Seven industrial countries to ascertain the effect of relevant fiscal rules. MULTIMOD, an empirically estimated macroeconomic world model subject to scrutiny both inside and outside the IMF, is well suited to analyze this question. The simulated rules are, in fact, stylized versions of the following: the proposed U.S. balanced-budget amendment; the Stability and Growth Pact under EMU, which calls for a fiscal target near balance, subject to a deficit limit of 3 percent of GDP and a debt limit of 60 percent of GDP; and the Japanese government‘s intention of adhering to its current balanced-budget rule in the future. Except for the U.S. proposal, all rules are assumed to allow for the operation of automatic stabilizers, under the specified constraints. This appendix consists of two parts: the first provides a description of the methodology and the second reports the simulation results.

Ordinarily, application of MULTIMOD for simulation analysis—in the context of the World Economic Outlook (WEO) exercises and Article IV consultations—is deterministic in that the residuals of the model are held constant. In the absence of policy shocks, the model tracks a baseline of endogenous variables. Such simulations are generally used to evaluate the effects of changes in a policy variable or those of exogenous shock on the values of output, exchange rates, interest rates, fiscal balance, inflation, or other endogenous variables. However, policymakers are interested not only in the short-run effects and dynamic repercussions of specified exogenous shocks but also in the variance caused by such shocks, including policy actions. From this perspective, stochastic simulations are particularly useful for comparing and evaluating various policy regimes. They can provide information on the robustness of a policy regime under plausible shocks and policy actions, consistent with historical experience.

In contrast to deterministic simulations, stochastic simulations require assumptions about the probability distribution of the residuals and, in some cases, of the exogenous variables. The probability distribution is generally derived from historical data over a sample period to produce a variance-covariance matrix of residuals or policy shocks or both. Repeated draws of residuals consistent with this probability distribution are used to solve the model. The results of these draws for the simulated endogenous variables are then evaluated by comparing key summary statistics (mean, mean-absolute error, root-mean-square error (RMSE), variance) across different policy rules.

Overall, the simulation results suggest that the largest output variability arises from strict adherence to a yearly balanced-budget rule through cuts in government consumption and is, on average, over 1 percent of GDP higher than in the baseline simulation. Not surprisingly, output variability is reduced significantly when the fiscal rule allows for the operation of automatic stabilizers. Moreover, implementation of the rule through adjustment in transfers or taxes displays a lower output variability than through adjustment in government consumption. However, operation of a flexible debt rule, if enforced, would not result in greater output variability than the baseline simulation. On balance, therefore, any increase in output variability may be outweighed by the credibility gains that ensue from fiscal rules.

Methodology

Stochastic simulations consist of repeated “trials,” or draws, from a standard normal distribution of random shocks, with a mean of 0 and a variance of 1, based on a simple transformation of a variance-covariance matrix.¹⁰⁵ This matrix is derived from the historical error terms of the model. For each draw, the model is simulated to produce the solution values for that trial. The process is repeated for successive periods until an arbitrary horizon is reached.¹⁰⁶ In this exercise, for each fiscal policy rule, the model was simulated 36 times for 10 years, for a total of 360 trials.¹⁰⁷ The simulated outcomes were then used to calculate the RMSE of key endogenous variables from their baseline paths.

The above procedure was subject to a number of modifications that affect the derivation of the variance-covariance matrix constructed with residuals consistent with the estimated equations of MULTIMOD. Whereas the first three modifications were similar to those used in an earlier paper (Masson and Symansky, 1992) and are only briefly mentioned, the treatment of fiscal policy, unique to this exercise, is discussed in detail.

First, given numerous data revisions and six additional years of observations since the original estimation of the model, some of the original equations may not fit as well as they did in 1990. Specifically, the historical residuals are no longer truly random (white noise) residuals. As it is beyond the scope of this exercise to reestimate the behavioral equations, the residuals have been whitened before computing the variance-covariance matrix. Autoregressive equations of the raw residuals—calculated from the behavioral equations—that include a constant, their own lagged value, and a time trend were estimated and added to the model.¹⁰⁸

Second, since expectations of certain variables appear in several of the estimated equations, proxies were formed using time-series techniques. The variables include human and capital wealth (which depend on expected future income) and expected future short-term interest rates, prices, and exchange rates.¹⁰⁹

Third, a specific assumption regarding exchange rate expectations—critical in the evaluation of alternative exchange rate and monetary regimes—was used for the interest parity equation. Although the residuals in the interest parity equation reflect a combination of default risk and expectation error, the model implicitly treats these residuals merely as a default premium. Although static expectation of exchange rates is somewhat inconsistent with the assumption used in other equations of the model, it is used in these simulations to reduce the size of the residuals.

Fourth, the fiscal data were redefined and equations were added to capture certain aspects of fiscal policy that are not part of the standard version of the model. In MULTIMOD, the tax variable is defined net of transfers under the assumption that transfers and taxes have the same effect on disposable income and economic activity. However, in this exercise, taxes and transfers were treated separately for several reasons: (a) taxes, especially capital and payroll taxes, tend to be more distortionary than transfers; (b) taxes tend to vary more over the cycle than transfers, the former displaying a higher elasticity with respect to income; and (c) their historical variability may exhibit significant differences. A separate equation for each fiscal variable was added to the model to incorporate their unexplained variation into the stochastic process, which represents the discretionary component of these policies. These include a government consumption equation that relates spending (as a ratio of potential output) to its lagged value, a constant, and a time trend; and similar equations for transfers and taxes, except that tax revenue is specified as a ratio of actual income. Furthermore, in the standard version of MULTIMOD, there is an intertemporal government budget constraint enforced through taxes that allows the automatic stabilizers to operate over the short term but does not allow debt to grow without limit. This equation was dropped from the model, but the nature of the stochastic shocks and the properties of the model ensure that this is not a problem in this exercise.¹¹⁰

Finally, the residuals of equations that were not critical to the analysis (such as the import of commodities) were dropped in order to reduce the magnitude of the exercise. Table 5, based on an estimation period of 1974–95, summarizes the properties of the autoregressive residual equations. The standard errors in these regressions characterize the main diagonal of the variance-covariance matrix.

Table 5.

Estimated Autoregression of Innovations, 1974–95

μ=α+ρμ_-1+βt+λt²+e

(Coefficient standard errors in parentheses)

Notes: μ is the raw residual, μ_-1 is the lagged residual, t is a time trend, and α,ρ,β, and λ are the estimated parameters; and σ_e is the standard error of the regression.

Table 5.

Estimated Autoregression of Innovations, 1974–95

μ=α+ρμ_-1+βt+λt²+e

(Coefficient standard errors in parentheses)

	United States			Japan			Germany			United Kingdom			Canada			France			Italy
	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe
Private consumption	0.39	(0.22)	0.010	0.07	(0.22)	0.016	0.32	(0.23)	0.023	0.43	(0.20)	0.016	0.10	(0.23)	0.018	0.31	(0.22)	0.008	0.32	(0.22)	0.012
Oil consumption	-0.02	0.23)	0.030	-0.34	(0.22)	0.038	0.33	(0.22)	0.044	-0.47	(0.21)	0.074	-0.42	(0.21)	0.052	-0.27	(0.23)	0.062	-0.49	(0.21)	0.043
Capital stock	0.70	(0.20)	0.007	0.35	(0.22)	0.010	0.64	(0.18)	0.005	0.08	(0.24)	0.016	0.58	(0.19)	0.006	0.55	(0.18)	0.004	-0.15	(0.23)	0.015
Manufactured exports	0.41	(0.21)	0.024	-0.11	(0.20)	0.039	0.26	(0.21)	0.039	0.33	(0.22)	0.019	0.40	(0.24)	0.033	0.60	(0.15)	0.023	0.09	(0.17)	0.028
Manufactured imports	0.06	(0.27)	0.031	0.35	(0.21)	0.066	0.05	(0.22)	0.028	-0.07	(0.24)	0.022	0.37	(0.22)	0.025	-0.05	(0.22)	0.019	-0.15	0.23)	0.022
Money	0.08	(0.22)	0.025	0.29	(0.16)	0.036	-0.09	(0.22)	0.055	0.02	(0.20)	0.077	0.35	(0.23)	0.025	-0.09	(0.22)	0.082	0.06	(0.22)	0.047
Long-term interest rate	0.63	(0.18)	0.012	0.48	(0.19)	0.008	0.54	(0.19)	0.010	0.34	(0.19)	0.012	0.64	(0.18)	0.012	0.65	(0.18)	0.012	0.62	(0.19)	0.013
GNP deflator	0.24	(0.22)	0.011	0.34	(0.22)	0.016	0.10	(0.24)	0.019	-0.14	(0.21)	0.019	0.39	(0.21)	0.011	0.18	(0.23)	0.008	0.40	(0.21)	0.009
Export deflator	0.16	(0.23)	0.028	0.04	(0.24)	0.027	-0.07	(0.24)	0.051	-0.16	(0.22)	0.020	0.20	(0.23)	0.034	0.19	(0.24)	0.015	0.33	(0.21)	0.023
Potential output	0.66	(0.18)	0.004	0.81	(0.16)	0.004	0.93	(0.08)	0.006	0.70	(0.17)	0.007	0.96	(0.16)	0.004	0.41	(0.18)	0.005	0.58	(0.15)	0.006
Government consumption	0.92	(0.99)	0.003	0.72	(0.17)	0.007	0.52	(0.19)	0.004	0.30	(0.19)	0.005	0.70	(0.19)	0.004	0.09	(0.19)	0.002	0.90	(0.10)	0.002
Tax revenue	0.42	(0.22)	0.004	0.59	(0.19)	0.010	0.56	(0.19)	0.015	0.50	(0.20)	0.011	0.52	(0.20)	0.010	0.62	(0.20)	0.008	0.58	(0.21)	0.030
Exchange rate	mdash;	mdash;	mdash;	0.36	(0.24)	0.024	0.84	(0.21)	0.015	0.46	(0.23)	0.020	0.39	(0.20)	0.011	0.39	(0.23)	0.018	0.30	(0.19)	0.024

Notes: μ is the raw residual, μ_-1 is the lagged residual, t is a time trend, and α,ρ,β, and λ are the estimated parameters; and σ_e is the standard error of the regression.

View Table

Table 5.

Estimated Autoregression of Innovations, 1974–95

μ=α+ρμ_-1+βt+λt²+e

(Coefficient standard errors in parentheses)

	United States			Japan			Germany			United Kingdom			Canada			France			Italy
	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe	ρ		σe
Private consumption	0.39	(0.22)	0.010	0.07	(0.22)	0.016	0.32	(0.23)	0.023	0.43	(0.20)	0.016	0.10	(0.23)	0.018	0.31	(0.22)	0.008	0.32	(0.22)	0.012
Oil consumption	-0.02	0.23)	0.030	-0.34	(0.22)	0.038	0.33	(0.22)	0.044	-0.47	(0.21)	0.074	-0.42	(0.21)	0.052	-0.27	(0.23)	0.062	-0.49	(0.21)	0.043
Capital stock	0.70	(0.20)	0.007	0.35	(0.22)	0.010	0.64	(0.18)	0.005	0.08	(0.24)	0.016	0.58	(0.19)	0.006	0.55	(0.18)	0.004	-0.15	(0.23)	0.015
Manufactured exports	0.41	(0.21)	0.024	-0.11	(0.20)	0.039	0.26	(0.21)	0.039	0.33	(0.22)	0.019	0.40	(0.24)	0.033	0.60	(0.15)	0.023	0.09	(0.17)	0.028
Manufactured imports	0.06	(0.27)	0.031	0.35	(0.21)	0.066	0.05	(0.22)	0.028	-0.07	(0.24)	0.022	0.37	(0.22)	0.025	-0.05	(0.22)	0.019	-0.15	0.23)	0.022
Money	0.08	(0.22)	0.025	0.29	(0.16)	0.036	-0.09	(0.22)	0.055	0.02	(0.20)	0.077	0.35	(0.23)	0.025	-0.09	(0.22)	0.082	0.06	(0.22)	0.047
Long-term interest rate	0.63	(0.18)	0.012	0.48	(0.19)	0.008	0.54	(0.19)	0.010	0.34	(0.19)	0.012	0.64	(0.18)	0.012	0.65	(0.18)	0.012	0.62	(0.19)	0.013
GNP deflator	0.24	(0.22)	0.011	0.34	(0.22)	0.016	0.10	(0.24)	0.019	-0.14	(0.21)	0.019	0.39	(0.21)	0.011	0.18	(0.23)	0.008	0.40	(0.21)	0.009
Export deflator	0.16	(0.23)	0.028	0.04	(0.24)	0.027	-0.07	(0.24)	0.051	-0.16	(0.22)	0.020	0.20	(0.23)	0.034	0.19	(0.24)	0.015	0.33	(0.21)	0.023
Potential output	0.66	(0.18)	0.004	0.81	(0.16)	0.004	0.93	(0.08)	0.006	0.70	(0.17)	0.007	0.96	(0.16)	0.004	0.41	(0.18)	0.005	0.58	(0.15)	0.006
Government consumption	0.92	(0.99)	0.003	0.72	(0.17)	0.007	0.52	(0.19)	0.004	0.30	(0.19)	0.005	0.70	(0.19)	0.004	0.09	(0.19)	0.002	0.90	(0.10)	0.002
Tax revenue	0.42	(0.22)	0.004	0.59	(0.19)	0.010	0.56	(0.19)	0.015	0.50	(0.20)	0.011	0.52	(0.20)	0.010	0.62	(0.20)	0.008	0.58	(0.21)	0.030
Exchange rate	mdash;	mdash;	mdash;	0.36	(0.24)	0.024	0.84	(0.21)	0.015	0.46	(0.23)	0.020	0.39	(0.20)	0.011	0.39	(0.23)	0.018	0.30	(0.19)	0.024

Notes: μ is the raw residual, μ_-1 is the lagged residual, t is a time trend, and α,ρ,β, and λ are the estimated parameters; and σ_e is the standard error of the regression.

Although interpretation of the estimation results is straightforward, it is difficult to disentangle the relative importance of the different variables. For example, the standard error of the investment equation is relatively small, but, as it represents unexplained variation in the capital stock (which is approximately three times the size of output), this stochastic element is numerically important. In contrast, the oil consumption equation, with a large standard error, has little effect on the variance of output because of the relatively small weight of oil.

The standard errors of the coefficients of the fiscal variables have potentially important implications. The unexplained component of government consumption is smaller than that of either taxes or transfers (Table 5), reflecting more inherent variability in the latter (Figure 1). This implies that fiscal rules that are met through adjustment in taxes or transfers may have relatively smaller effects on output since this adjustment may offset the initial source of variance. Similarly, since government consumption has been historically stable, reliance on this variable to balance the budget is likely to add to output variability.

General Government Revenue and Expenditure

Source: International Monetary Fund, World Economic Outlook database.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

General Government Revenue and Expenditure

Source: International Monetary Fund, World Economic Outlook database.

• Download Figure
• Download figure as PowerPoint slide

General Government Revenue and Expenditure

Source: International Monetary Fund, World Economic Outlook database.

• Download Figure
• Download figure as PowerPoint slide

Baseline Simulation

The baseline simulation serves as a standard of comparison for the simulation of selected fiscal rules. It includes the fiscal equations, discussed above, that allow for the operation of automatic stabilizers as well as discretionary policy through the stochastic elements. As previously indicated, government consumption and transfer payments grow as a function of potential output and tax revenues of actual income. Whether fiscal rules increase or decrease total variability will depend on the relative size and persistence of exogenous fiscal policy shocks, compared with those of the other underlying shocks, and on their interplay with automatic stabilizers.

Comparison of the variability in the baseline simulation and the actual historical variability of endogenous variables¹¹¹ reveals that the baseline simulation represents a reasonable, although not perfect, characterization of macroeconomic variability (Table 6). Thus, the baseline can justifiably be used to assess the effects of alternative fiscal rules. In general, GDP is used as the summary variable to compare the similarity of the historical variability and the baseline variability. GDP variation differs by nearly 0.5 percentage point for the United States and France and 0.7 percentage point for Italy, and is identical to its historical variability for Canada. In the case of Canada, however, GDP variability masks some more significant differences in the variability of interest rates and investment. For Germany and the United Kingdom, the baseline output variability is over 1 percentage point higher than its historical variability. Meanwhile, the baseline variability of output in Japan is 1.5 percentage points greater than its historical variability—apparently as a result of the poor tracking performance of investment in MULTIMOD. Somewhat surprisingly, the baseline in Japan shows smaller variation in the exchange rate and government deficits. One way of putting in perspective the magnitude of the difference between the historical and baseline variance is to compare them with their forecast errors. An analysis of the forecast errors for major industrial countries (IMF, 1996a, Annex I) revealed that for real GDP growth in industrial countries the average current-year RMSE was 0.72, and 1.46 for the one-step-ahead forecast, with the smallest errors for the United States and France and the largest for Germany, Italy, and Japan. These results are consistent with the differences between the historical and baseline output variations.

Table 6.

Variability of Endogenous Variables Under Alternative Fiscal Rules

Note: LC = local currency.

Table 6.

Variability of Endogenous Variables Under Alternative Fiscal Rules

(Root-mean-square error, in percentage terms)
			Policy Simulations
			Balanced-budget target
	Actual (1974–95)	Baseline Simulation	Government consumption adjustment (1)	Transfer adjustment (2)	Tax adjustment (3)	No deficit (4)	Deficit ceiling (3 percent of GDP)(5)	Debt target (6)
United States
GDP	2.12	1.78	2.60	1.89	1.92	1.97	1.83	1.82
Long-term interest rate	2.21	1.40	1.16	1.34	1.32	1.33	1.34	1.22
Inflation	2.42	1.28	1.58	1.31	1.30	1.45	1.35	1.38
Exchange rate ($/LC)	–	–	–	–	–	–	–	–
Government consumption	0.87	0.53	1.33	0.55	0.55	1.48	0.64	0.71
Tax revenue	0.59	0.45	0.47	0.45	1.41	0.50	0.45	0.44
Transfers	1.00	0.57	0.73	1.17	0.61	0.63	0.59	0.60
Fiscal balance	1.52	1.57	–	–	–	1.61	1.26	1.01
Government debt	11.39	6.61	2.13	1.53	1.55	7.32	5.26	3.71
Japan
GDP	1.61	3.29	5.22	3.58	3.61	3.85	3.58	3.21
Long-term interest rate	1.91	1.25	1.24	1.21	1.18	1.43	1.20	1.17
Inflation	2.30	2.30	2.84	2.35	2.35	2.54	2.43	2.32
Exchange rate ($/LC)	11.17	9.45	7.54	8.91	8.42	8.59	8.48	7.40
Government consumption	1.45	1.01	2.00	1.06	1.06	1.62	0.83	0.82
Tax revenue	3.52	1.11	1.10	1.09	2.20	1.14	1.10	1.09
Transfers	1.60	1.45	1.87	1.65	1.53	1.59	1.52	1.47
Fiscal balance	3.08	2.11	–	–	–	2.02	1.70	1.81
Government debt	9.56	6.51	1.54	1.09	1.10	8.50	4.76	4.56
Germany
GDP	1.93	3.06	3.93	3.16	3.13	3.30	3.13	3.04
Long-term interest rate	1.22	1.98	1.79	1.91	1.87	2.11	1.99	1.98
Inflation	1.17	1.52	1.89	1.55	1.55	1.70	1.61	1.63
Exchange rate ($/LC)	11.41	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	1.60	0.65	2.67	0.68	0.68	1.66	1.04	0.96
Tax revenue	2.32	1.52	1.59	1.53	2.31	1.58	1.52	1.51
Transfers	1.77	0.98	1.60	1.96	1.15	1.26	1.13	1.15
Fiscal balance	2.79	2.37	–	–	–	2.04	1.95	1.75
Government debt	14.47	8.83	2.89	2.08	2.10	9.32	7.45	6.09
Canada
GDP	2.32	2.34	3.02	2.44	2.45	2.46	2.35	2.35
Long-term interest rate	2.05	1.64	1.71	1.64	1.62	1.81	1.67	1.72
Inflation	3.18	1.78	1.99	1.79	1.79	1.88	1.82	1.80
Exchange rate ($/LC)	4.13	8.21	7.64	8.11	8.04	7.97	8.02	7.68
Government consumption	1.18	0.65	1.82	0.67	0.67	1.25	0.55	0.80
Tax revenue	2.53	0.96	0.94	0.93	1.90	0.96	0.95	0.93
Transfers	1.71	1.01	1.24	1.22	1.06	1.09	1.05	1.05
Fiscal balance	2.37	1.70	–	–	–	1.55	1.42	1.19
Government debt	21.27	6.79	2.84	2.47	2.48	7.45	5.90	4.22
France
GDP	1.44	1.97	3.33	2.15	2.19	2.41	2.09	2.08
Long-term interest rate	2.53	1.47	1.30	1.40	1.36	1.53	1.47	1.36
Inflation	3.81	2.09	3.33	2.25	2.22	2.47	2.17	2.36
Exchange rate ($/LC)	12.29	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	0.79	0.44	1.91	0.49	0.50	1.03	0.42	0.41
Tax revenue	2.77	0.92	0.91	0.88	1.78	0.87	0.88	0.88
Transfers	2.08	0.89	1.37	1.19	0.98	1.06	0.93	0.96
Fiscal balance	2.33	1.64	–	–	–	1.37	1.49	1.33
Government debt	10.08	6.34	4.02	2.93	2.81	6.72	6.03	5.53
Italy
GDP	2.07	2.76	4.51	2.99	3.00	3.35	3.13	3.30
Long-term interest rate	2.83	2.14	1.99	2.08	2.05	2.19	2.14	2.02
Inflation	5.20	2.55	4.40	2.73	2.72	3.35	3.10	3.80
Exchange rate ($/LC)	12.84	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	0.94	0.55	4.79	0.57	0.58	3.81	2.77	3.50
Tax revenue	5.02	3.58	3.51	3.61	3.60	3.39	3.51	3.52
Transfers	1.34	1.69	2.16	3.55	1.79	1.91	1.87	1.96
Fiscal balance	3.70	4.00	–	–	–	4.64	3.89	2.03
Government debt	22.23	16.16	11.92	7.53	7.29	19.20	16.95	13.20
United Kingdom
GDP	2.11	3.33	5.94	3.78	3.81	4.44	3.93	3.69
Long-term interest rate	2.17	1.78	1.47	1.68	1.63	1.74	1.70	1.66
Inflation	5.57	3.34	5.21	3.59	3.57	4.19	3.72	3.92
Exchange rate ($/LC)	10.26	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	2.77	0.91	2.59	1.02	1.02	1.72	1.02	0.71
Tax revenue	1.83	1.21	1.14	1.16	2.48	1.17	1.18	1.17
Transfers	1.18	1.17	1.69	1.83	1.27	1.44	1.34	1.26
Fiscal balance	3.13	2.44	–	–	–	1.87	1.91	2.04
Government debt	8.49	8.37	5.21	3.59	3.50	8.86	7.38	7.39

Note: LC = local currency.

View Table

Table 6.

Variability of Endogenous Variables Under Alternative Fiscal Rules

(Root-mean-square error, in percentage terms)
			Policy Simulations
			Balanced-budget target
	Actual (1974–95)	Baseline Simulation	Government consumption adjustment (1)	Transfer adjustment (2)	Tax adjustment (3)	No deficit (4)	Deficit ceiling (3 percent of GDP)(5)	Debt target (6)
United States
GDP	2.12	1.78	2.60	1.89	1.92	1.97	1.83	1.82
Long-term interest rate	2.21	1.40	1.16	1.34	1.32	1.33	1.34	1.22
Inflation	2.42	1.28	1.58	1.31	1.30	1.45	1.35	1.38
Exchange rate ($/LC)	–	–	–	–	–	–	–	–
Government consumption	0.87	0.53	1.33	0.55	0.55	1.48	0.64	0.71
Tax revenue	0.59	0.45	0.47	0.45	1.41	0.50	0.45	0.44
Transfers	1.00	0.57	0.73	1.17	0.61	0.63	0.59	0.60
Fiscal balance	1.52	1.57	–	–	–	1.61	1.26	1.01
Government debt	11.39	6.61	2.13	1.53	1.55	7.32	5.26	3.71
Japan
GDP	1.61	3.29	5.22	3.58	3.61	3.85	3.58	3.21
Long-term interest rate	1.91	1.25	1.24	1.21	1.18	1.43	1.20	1.17
Inflation	2.30	2.30	2.84	2.35	2.35	2.54	2.43	2.32
Exchange rate ($/LC)	11.17	9.45	7.54	8.91	8.42	8.59	8.48	7.40
Government consumption	1.45	1.01	2.00	1.06	1.06	1.62	0.83	0.82
Tax revenue	3.52	1.11	1.10	1.09	2.20	1.14	1.10	1.09
Transfers	1.60	1.45	1.87	1.65	1.53	1.59	1.52	1.47
Fiscal balance	3.08	2.11	–	–	–	2.02	1.70	1.81
Government debt	9.56	6.51	1.54	1.09	1.10	8.50	4.76	4.56
Germany
GDP	1.93	3.06	3.93	3.16	3.13	3.30	3.13	3.04
Long-term interest rate	1.22	1.98	1.79	1.91	1.87	2.11	1.99	1.98
Inflation	1.17	1.52	1.89	1.55	1.55	1.70	1.61	1.63
Exchange rate ($/LC)	11.41	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	1.60	0.65	2.67	0.68	0.68	1.66	1.04	0.96
Tax revenue	2.32	1.52	1.59	1.53	2.31	1.58	1.52	1.51
Transfers	1.77	0.98	1.60	1.96	1.15	1.26	1.13	1.15
Fiscal balance	2.79	2.37	–	–	–	2.04	1.95	1.75
Government debt	14.47	8.83	2.89	2.08	2.10	9.32	7.45	6.09
Canada
GDP	2.32	2.34	3.02	2.44	2.45	2.46	2.35	2.35
Long-term interest rate	2.05	1.64	1.71	1.64	1.62	1.81	1.67	1.72
Inflation	3.18	1.78	1.99	1.79	1.79	1.88	1.82	1.80
Exchange rate ($/LC)	4.13	8.21	7.64	8.11	8.04	7.97	8.02	7.68
Government consumption	1.18	0.65	1.82	0.67	0.67	1.25	0.55	0.80
Tax revenue	2.53	0.96	0.94	0.93	1.90	0.96	0.95	0.93
Transfers	1.71	1.01	1.24	1.22	1.06	1.09	1.05	1.05
Fiscal balance	2.37	1.70	–	–	–	1.55	1.42	1.19
Government debt	21.27	6.79	2.84	2.47	2.48	7.45	5.90	4.22
France
GDP	1.44	1.97	3.33	2.15	2.19	2.41	2.09	2.08
Long-term interest rate	2.53	1.47	1.30	1.40	1.36	1.53	1.47	1.36
Inflation	3.81	2.09	3.33	2.25	2.22	2.47	2.17	2.36
Exchange rate ($/LC)	12.29	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	0.79	0.44	1.91	0.49	0.50	1.03	0.42	0.41
Tax revenue	2.77	0.92	0.91	0.88	1.78	0.87	0.88	0.88
Transfers	2.08	0.89	1.37	1.19	0.98	1.06	0.93	0.96
Fiscal balance	2.33	1.64	–	–	–	1.37	1.49	1.33
Government debt	10.08	6.34	4.02	2.93	2.81	6.72	6.03	5.53
Italy
GDP	2.07	2.76	4.51	2.99	3.00	3.35	3.13	3.30
Long-term interest rate	2.83	2.14	1.99	2.08	2.05	2.19	2.14	2.02
Inflation	5.20	2.55	4.40	2.73	2.72	3.35	3.10	3.80
Exchange rate ($/LC)	12.84	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	0.94	0.55	4.79	0.57	0.58	3.81	2.77	3.50
Tax revenue	5.02	3.58	3.51	3.61	3.60	3.39	3.51	3.52
Transfers	1.34	1.69	2.16	3.55	1.79	1.91	1.87	1.96
Fiscal balance	3.70	4.00	–	–	–	4.64	3.89	2.03
Government debt	22.23	16.16	11.92	7.53	7.29	19.20	16.95	13.20
United Kingdom
GDP	2.11	3.33	5.94	3.78	3.81	4.44	3.93	3.69
Long-term interest rate	2.17	1.78	1.47	1.68	1.63	1.74	1.70	1.66
Inflation	5.57	3.34	5.21	3.59	3.57	4.19	3.72	3.92
Exchange rate ($/LC)	10.26	8.78	7.09	8.43	7.63	7.62	8.09	7.32
Government consumption	2.77	0.91	2.59	1.02	1.02	1.72	1.02	0.71
Tax revenue	1.83	1.21	1.14	1.16	2.48	1.17	1.18	1.17
Transfers	1.18	1.17	1.69	1.83	1.27	1.44	1.34	1.26
Fiscal balance	3.13	2.44	–	–	–	1.87	1.91	2.04
Government debt	8.49	8.37	5.21	3.59	3.50	8.86	7.38	7.39

Note: LC = local currency.

The historical variation in the government deficit ranges from 1.5 percent of GDP for the United States to 3.7 percent of GDP for Italy. Allowing the automatic stabilizers to work, as in the baseline simulation, results in a deficit variation similar to the historical variation for these two countries. Again, Japan tends to show the most sizable difference in deficit variation: 3 percent of GDP historically, but only 2 percent of GDP when automatic stabilizers are allowed to operate.

As noted earlier, taxes and transfers tend to show substantially more variation than government consumption, and, with the exception of the United States, taxes are more variable than transfers. On the other hand, even in cases where taxes historically have been more variable than transfers, they are not necessarily more unpredictable, as seen in the baseline (Japan and Canada); meanwhile, government consumption remains less variable than the other two fiscal variables.

Comparison of Fiscal Rules

Six policy simulations were performed: (1) balanced budget enforced through government consumption adjustment; (2) balanced budget enforced through transfer adjustment; (3) balanced budget enforced through tax adjustment; (4) no-deficit rule, but allowance for automatic stabilizers when in surplus; (5) a 3 percent of GDP budget deficit ceiling with allowance for automatic stabilizers below the ceiling; and (6) debt target, with allowance for short-run debt accumulation or drawdown. All the rules, except (2) and (3), are assumed to be met through adjustment in government consumption. The rules are stylized in that they are binding and are always realized ex post. However, rules (4) through (6) allow for the operation of automatic stabilizers, namely, downward in (4) and symmetrically in (6), whereas (5) is symmetric up to the ceiling. The debt rule (6) represents the closest approximation to the “tax-smoothing” function of government. Although it allows the automatic stabilizers to work, this rule also attempts to offset the debt buildup that can occur over the cycle and effectively adds a stock condition to the debt target rule. This is analogous to the discussion in the monetary literature, where it is recognized that an inflation rule (deficit) allows for “drift” in the price level (debt).

There is no metric used in this paper to compare the various regimes. Conceptually, a welfare or utility measure could be used, but the results would depend upon subjective weighting of the mean and variance of different endogenous and policy variables. Rather, Table 6 lets the reader compare the various regimes. The discussion below concentrates on output variability because it is associated with some important costs, including, for example, macroeconomic uncertainty and permanent unemployment if the labor market can be characterized with hysteresis.¹¹² To put the RMSE comparisons in perspective, relative to the baseline simulation, for example, the 0.8 percentage point increase in output variability in the United States under simulation (1) may be viewed as large since it represents a 30 percent increase in variability. On the other hand, this difference could be considered small if compared with the average RMSE of WEO forecasts of GDP for the major industrial countries.

Several general conclusions can be drawn from the results (Table 6). First, relative to the baseline with the automatic stabilizers operating, when tight deficit control is implemented through adjustment of government consumption, in simulation (1) it tends to raise the variability of output, on average, by 1.4 percentage points, representing a 50 percent increase. This rule increases output variability from a low of just under 0.7 percentage point in Canada, and 0.85 in the United States and Germany, to substantially over 1 percentage point in France, and around 2 percentage points in Italy, Japan, and the United Kingdom. This increase in variability is the result of not allowing the automatic stabilizers to operate, as well as of following a procyclical policy, which more than outweighs the variability introduced through the stochastic elements of discretionary policy. However, the countries that exhibited the smallest increase in output variation generally have a more stable fiscal policy in the baseline (Canada, France, and United States), which limits the need for fiscal offset. Prices, which tend to be procyclical in the baseline, also exhibited more variability under the strict balanced-budget rule. Additionally, with government saving less variable, long-term interest rates were more stable under rule (1) than in the baseline.

The no-deficit rule, under simulation (4), which represents an asymmetric implementation of a strict expenditure-enforced deficit target, exhibits reduced variation relative to the comparable symmetric rule in (1). Although, on average, this simulation shows the second highest output variability, it is only, on average, 0.5 percentage point greater than in the baseline and less than 0.25 for over half the countries. This implies that a balanced-budget rule, if realistically implemented in an asymmetric fashion, might add little additional variability to output while providing a surplus bias to fiscal policy.¹¹³

The debt rule, given in (6), combines the tax-smoothing role of fiscal policy with long-term debt reduction and control. The automatic stabilizers are allowed to work over the short run as long as the stock of public debt remains relatively stable. This represents a pragmatic rule in that the speed at which a country must achieve a debt target is not precisely given. By letting the automatic stabilizers work, this rule offsets demand fluctuations, but, with one eye on debt control, it also avoids irresponsible fiscal policy. Under this fiscal rule, output variability is nearly identical to the baseline for most countries, with two countries showing reduced variability. Only in the case of Italy does debt stabilization show significantly more variation than when the automatic stabilizers are allowed to operate freely. This is largely owing to the high debt stock, which makes debt stabilization difficult over the short run.¹¹⁴

A comparison of fiscal rules based on the instrument of compliance suggests that adjustment in taxation or transfer payments tends to produce less variance than adjustment in government consumption. This appears to be inconsistent with the findings of Alesina and Perotti (1995) and McDermott and Wescott (1996), who conclude that successful fiscal policy is achieved through expenditure control. These results can be viewed, however, as providing partial support to those findings. Historically, tax ratios tend to show more variation than government consumption ratios (Figure 1). Therefore, if variation in the tax burden is an explanation for the variability of deficits, then achievement of a balanced budget through tax adjustment would remove a source of macroeconomic variability. Although tax variability in (3) is greater than in the baseline, in most countries it shows a significantly smaller increase in variation than government consumption did when the deficit target was achieved with spending adjustment. For Italy, in particular, tax variation is almost identical to the baseline and output variability is similar, implying that deficit variation is due to both discretionary tax policy and the automatic stabilizers. Only for the United States does the amount of tax variation increase significantly, and this is more a reflection of a relatively stable tax ratio in the baseline.

The general finding that output variability under tax or transfer adjustment is only modestly different than in the baseline partly reflects the undoing of a variable tax policy but also reflects the moderate Ricardian properties of MULTIMOD, indicating that tax changes have more muted effects on output than changes in government consumption. When deficit targets are achieved through changes in transfer payments, the results are remarkably similar to the regime where deficit targets are achieved through taxes. This partly reflects the similarity in tax and transfer variability in the baseline. In countries where taxes in the baseline show more variability than transfers (Germany), output variability tends to increase with transfer adjustment. The opposite occurs when transfers are more variable (the United States); when they are similar (Canada), the differences in these regimes are minimal. Additionally, relying on transfers rather than taxes tends to reduce output variation since transfer payments usually have less distortionary effects on private behavior than tax changes do.

The results of the deficit ceiling simulation (5) are similar to the baseline simulations, as the former can be regarded as a balanced-budget target, allowing for automatic stabilizers to operate symmetrically. Only in the case where the deficit target is breached is a strict fiscal reaction required and thus do the results show some reduction in the variance of debt. Given the variance in output in the baseline case, there are few occasions when the fiscal authorities would have to offset the automatic stabilizers to achieve the target.¹¹⁵

Finally, monetary policy is largely ignored in the analysis. The baseline assumes that exchange rates are fixed within EMU and that other countries target a stable money growth path. However, to the extent that fiscal policy has constrained the flexibility of monetary policy, moving to a fiscal rule has the potential to give increasing flexibility to monetary policy to stabilize output. Analysis of this important issue, nonetheless, is beyond the scope of this paper.