Chapter

Chapter 6. Public Debt Targeting: An Application to Caribbean Countries

Author(s):
R. Gelos
Published Date:
March 2014
Share
  • ShareShare
Show Summary Details
Author(s)
Alejandro Guerson and Giovanni Melina 

It is a well-established fact in the international finance literature that developments in international financial markets can be a source of instability, and can amplify real shocks (see, for example, Kaminsky, Reinhart, and Vegh, 2004). The channels include the direct impact of capital flows on aggregate demand, including through domestic financial markets, and also the multiplication effect from the impact of capital flows on government access to financing and borrowing costs. In countries with high public debt, negative real or financial shocks can undermine creditworthiness, perhaps to the point at which the sovereign’s ability to honor its debt obligations is perceived to be at risk. At that point, a second round of indirect effects can sink an economy into a negative spiral of high interest rates, credit crunch, and recession.

According to the logic above, the stock of public debt is a critical state variable of the economy, not just in terms of determining its sustainable level of consumption and public sector spending in the long term, but also in terms of affecting the business cycle dynamics in the short term. A sovereign debtor subject to financial markets’ swings might (1) be forced into procyclical expenditures, if access to financing becomes too costly or just unavailable in low states of the cycle, further amplifying business cycle fluctuations, and (2) need additional self-insurance against loss of access to financing, as opposed to using public debt and access to financing as a source of insurance against macroeconomic shocks. Moreover, the financial contagion literature indicates that the negative dynamics described above can take place even if direct channels and change in fundamentals do not necessarily warrant the financial markets’ reaction outlined (see, for example, Kaminsky and Reinhart, 2000). Points (1) and (2) above imply that keeping public debt at sufficiently low levels is a form of self-insurance, as it allows expenditure smoothing over the business cycle by permitting access to financing without jeopardizing creditors’ perceptions about a sovereign’s ability to honor debt commitments.

This chapter proposes a fiscal framework that, based on the reasoning above, seeks to smooth primary expenditures over the business cycle by providing financial creditors sufficient assurance of a high probability that public debt obligations will remain within a preannounced band. The use of a public debt band is critical, as it allows for primary balance flexibility and, as a result, avoids the need for large fiscal consolidations in bad states of the cycle, which would undermine credibility. Operationally, the fiscal framework proposal consists of the announcement of four intermediate parameters that pin down the primary expenditure envelope for the next budget cycle, which is set so that public debt remains below a preannounced threshold in the medium term with a specified probability. Empirical simulations for a sample of Caribbean countries show that allowing debt to move within a relatively narrow band allows for large improvements in terms of primary expenditure smoothing in the business cycle. The simulations in this chapter indicate that expenditure smoothing can be better than under a structural balance rule. This is a tough test for the proposal, as the Caribbean economies are typically highly indebted, very open, undiversified, and subject to significant real and financial shocks.

The proposed framework has the following desirable features: (1) it balances the direct impact of countercyclical government spending on aggregate demand vis-à-vis the indirect feedback effect of excess public sector deficit and public debt accumulation, as these can affect sovereign risk and interest rates; (2) it can enhance credibility as reputation builds over time; (3) it does not need a potential output estimation for the current year; (4) it is state-contingent, therefore permitting some overall balance flexibility; and (5) it forces the use of a consistent medium-term macroeconomic framework for budgeting. As anticipated, the framework is applied to Caribbean economies, which, in general show high levels of public debt and procyclical primary expenditure, as will be shown in the section on Application to Caribbean Countries. Details on the computation of debt dynamics are provided in Appendix 6.1.

The Proposal: Public Debt Targeting

We propose a fiscal framework under which a government commits to a high probability of keeping debt within a specified band. The policy entails two main parameter announcements: a trajectory of public debt and its probabilistic band, and a primary expenditure budget envelope for the next budget cycle. The two main announcements are linked by four intermediate technical parameters, which are also announced: (1) a medium-term horizon in years; (2) a public debt upper threshold at the end of the medium-term horizon; (3) a notional spending trajectory that maps the next year’s primary expenditure announcements to debt trajectories over the medium term; and (4) a probability value attached to debt remaining below the upper threshold.

Operationally, under the public debt targeting framework, a government is asked to set primary spending budget envelopes for the next year according to probabilistic public debt outcomes over the medium term. Concretely, in year t, a government would announce a primary spending budget for the year, S(t) (budget envelope), such that if primary spending for the following T years (medium-term horizon) were to remain at the same level in real per capita terms (spending trajectory), public debt would take a level D(t+T) percent of GDP or lower (debt upper threshold) with probability P(t+T) (probability). In year t+1, the government would repeat the same announcement, choosing a level of spending S(t+1) such that after T years, public debt takes at most a level of D(t+T+1) with probability P(t+T+1), and so on.

The proposed framework is recursive in nature. Budgets are adjusted every year for a given choice of targets and thresholds after the impact of shocks on the previous year’s deficit and public debt has been observed. For example, if a negative shock results in lower revenues in year t compared with the budget projection, then, for a given S(t) budget, the deficit and the amount of debt accumulation would be higher than originally expected. All else being equal, this would result in a lower S(t+1) budget envelope relative to the one that would have been determined by the framework without the negative shock, for a given set of medium-term debt target threshold, probability value, and horizon length. In this way, the proposal explicitly tackles the issue of the trade-off between expenditure smoothing over the business cycle and the need to avoid sharp increases in public debt, as the need for expenditure adjustment is averaged away over a sequence of periods.

The proposal can, more generally, be expanded to accommodate a preannounced degree of discretionary expenditure. The spending trajectory can be set up in terms of an acyclical notional trajectory, as in the example above, or expanded with additional procyclical or countercyclical impulses around the primary spending level in real per capita terms. The simulations to be presented below show an example in this regard.1 In the example, we simulate a discretionary spending space such that, if real GDP growth is higher (lower) than population growth, a specified share can be spent (saved) in the current budget period. This discretionary stimulus or expenditure impulse can be set to be countercyclical or procyclical, as deemed appropriate.2

The proposal has several desirable properties in terms of fiscal policy management:

Credibility

Commitment to a public debt band is anchored on a specific framework that sets government spending according to projected debt trajectories—this is a critical feature in terms of building credibility. For example, negative shocks that reduce GDP and revenues and result in a larger deficit and public debt relative to budget projections would be less likely to trigger expectations of public debt entering an unsustainable path, as a strategy for a compensating consolidation is automatically in place. Budget announcements acknowledge up-front the possibility of shocks and the resulting deviations, but at the same time they are clear about the mechanism by which public debt will be brought back on track. As a government abides by its policy framework announcements, reputation can build up, making access to financing in bad times more fluid and on better terms. In addition, the framework proposal sets primary spending on a smooth trajectory, which also contributes to increasing the credibility of the debt band announcement. This is because it significantly reduces the need for unrealistically large primary spending consolidation in bad times, as government deficits and debt accumulation are averaged away over a sequence of budget cycles.

Primary Expenditure Smoothing

Because debt thresholds are set according to medium-term economic paths, deficits can be larger than budgeted if revenues turn out lower than projected while primary expenditures remain at the budget level. The focus on primary spending, as opposed to total expenditures, is not a shortcoming of public debt targeting. First, for a given tax framework in place, primary expenditures are the relevant expenditure aggregate in terms of assessing debt sustainability. And second, interest expenditures tend to be largely predetermined. Excluding interest expenditures from the spending announcement implies that shocks to interest costs (for example, from interest rates or exchange rates) are accommodated.

Flexibility

The proposed framework is not a spending rule. Under public debt targeting, the government is not committing to keep primary spending constant in real per capita terms. That is a notional spending trajectory used as an intermediate technical parameter to pin down primary expenditures for the next budget, jointly with the other intermediate technical parameters outlined above. Each year, primary spending is recalculated recursively, as explained above. The framework is set in terms of primary expenditures, which implies that automatic stabilizers on revenues are allowed to play in full, while unanticipated changes in interest expenditures are accommodated.

Medium-Term Macroeconomic Framework

Public debt targeting requires fiscal projections over the medium term, as determined by its various parameter choices. Probabilistic projections need to be determined according to a relatively agnostic representation of the economy (a “model”), which is fully disclosed and can therefore be subject to scrutiny. This implies that public debt targeting forces medium-term budget planning. Institutional mechanisms to ensure independence from political influence and stability of the representation are likely needed. At a minimum, the framework requires minimizing the space for arbitrary and unreasonable macroeconomic parameters in the budget and in medium-term projections.

A key input is therefore an economic model that can be used to produce consistent macroeconomic forecasts incorporating the possibility of shocks. The model should be capable of providing stochastic forecasts for the variables relevant to make probabilistic public debt projections, although other variables considered important determinants of the business cycle should also be included.

Public Debt Targeting and structural Balance Rules

Kopits and Symansky (1998) define fiscal rules as numerical constraints set on the budget on a permanent basis. If we take this definition, then the proposal in this chapter is not a rule: what is stable over time is the framework, as determined by the choices of the four intermediate medium-term parameters involved and the economic model of choice, but not the actual numerical targets. Under public debt targeting, primary expenditure and debt levels change in every budget cycle, as the budget envelope for the following year is recalculated after observing the impact of shocks on public debt.3

Besides definitional issues, the framework compares favorably with some alternatives. Numerical constraints, particularly on the fiscal balance—as in the Maastricht Treaty signed by the European Union—have the advantage of simplicity. This could be argued as being important if public scrutiny is considered necessary when special interests or other political economy pressures for spending are strong. However, by their nature, such constraints would not necessarily suffice to avoid procyclical spending or excessive debt accumulation—the two main objectives typically sought.4

This is why proponents of fiscal rules prefer to focus on the structural or cyclically adjusted balance rules. These are based on measures of the structural or noncyclical components of revenues and expenditures, and therefore account for the transitory impact of business cycle fluctuations on the budget and other transitory or one-off factors. By their nature, provided structural balance targets are set at an appropriate level, these should be consistent with both debt sustainability and a nonprocyclical stance. However, there are disadvantages or limitations often attached to these rules: (1) difficulty in discriminating permanent versus transitory changes in revenues and expenditures, which can undermine transparency and credibility; (2) insufficiency of GDP to assess in full the state of the cycle; and (3) inability to allow space for discretionary fiscal policy.

These limitations are not necessarily fair, as refinements can be introduced to improve structural balance rules’ application in practice.5 Regardless, public debt targeting would not have these limitations. First, the identification of permanent and transitory output components is not necessary.6 Instead, spending levels are determined based on projected public debt dynamics, which is simpler and more transparent as these are anchored around the debt accumulation identity. Second, given the wider set of indicators needed to produce debt projections, the methodology forces the use of a broader set of variables to determine appropriate primary spending levels, which are also crucial in determining the state of the business cycle, such as interest rates and real exchange rates.7 Third, the proposed framework tackles the impact of debt sustainability concerns on the business cycle in the most direct way possible, given the commitment to public debt thresholds and the announcement of a debt trajectory band. Fourth, as explained above, the framework can be designed to allow for cyclical space for discretionary primary spending, possibly including a state-contingent additional primary expenditure component (set to be either procyclical or countercyclical, as deemed appropriate), but without necessarily allowing full discretion.

In our view, however, the most promising application of public debt targeting is as a transition framework toward the adoption of a structural balance rule when public debt is high and marginal deficits and public debt accumulation can impact sovereign spreads. Expenditure smoothing over the business cycle requires the ability to finance fiscal deficits, and therefore relies on the assumption that markets would tolerate deterioration in creditworthiness in bad times. To the extent that public debt targeting is an intermediate step that seeks to reduce debt within certain bounds consistent also with expenditure smoothing, it can then be a suitable framework for a transition toward lower indebtedness. The use of a debt band, as it results from the framework proposal in this chapter, balances out the objective to smooth primary expenditures over the business cycle with the need to avoid sharp increases in public debt.8

Application to Caribbean Countries

The Caribbean economies are good candidates for our proposal in this chapter, given their high public debt levels and procyclical primary expenditure. The empirical evidence presented below is indicative of financial developments associated with this behavior.

The Conduct of Fiscal Policy in Caribbean Countries: Some Stylized Facts

We start by establishing some stylized facts that are indicative of the relevance of the policy proposal for Caribbean countries.

High public-debt-to-GDP ratios. Since 2009, the median public-debt-to-GDP ratio has remained at about 90 percent of GDP, and about four times the size of total revenues (Figure 6.1, panel a). Most countries show very high public debt ratios, and also in almost all cases, debt-to-GDP ratios are increasing or persistently high (Figure 6.1, panel b).

Figure 6.1Fiscal Behavior and the Business Cycle

Source: IMF staff calculations.

Procyclical primary expenditures. The median cyclical components of real output, primary expenditures, and total revenues show positive co-movement over the last 15 years (Figure 6.1, panel c). The cyclical components’ median correlation of primary expenditures with both total revenues and real GDP is about 50 percent. This is a general pattern across the developing world.

Strong cross-country synchronization of GDP, government revenues, and primary expenditures, suggesting that common external factors are important. Virtually all countries show a positive correlation of the output, total revenue, and primary expenditure cycles with cross-country median fluctuations (Figure 6.1, panel d).

Strong cross-country correlation of domestic financial and foreign exchange markets, possibly corresponding to the state of the U.S. business cycle. Every country’s GDP cycle is positively correlated with that of the United States; there is also evidence of cross-country correlation in interest rates, real exchange rates, and net capital inflows (Figure 6.1, panel e). This suggests that external financial developments are an important channel of transmission of external real and financial shocks.

Financial variables explain a substantial share of the variation in primary expenditures. Net capital inflows, real exchange rate depreciation, and interest rates jointly explain one-half of the average forecast-error variance of primary expenditures (Figure 6.1, panel f).9 The statement assumes that a significant share of variation in financial indicators has an external (exogenous) source, as suggested by the real and financial cross-country correlations.

Simulation

The starting point of the simulation is the public debt accumulation identity. We decompose public debt to GDP into: (1) the commercial component denominated in the domestic currency; (2) the commercial component denominated in foreign currency;10 and (3) the official bilateral/multilateral fraction of public debt (usually denominated in U.S. dollars). Let dtj be the annual public-debt-to-GDP component of type j (for example, the component denominated in local currency). dtj evolves according to the following:

where itj,gt,πt,αtj and bt represent the interest rate paid on a government debt of type j, the real GDP growth rate, the annual inflation rate, the share of primary balance servicing public debt of type j, and the primary-balance-to-GDP ratio, respectively. bt is given by the difference between total revenues and primary (non-interest) expenditures divided by GDP. The sum of all types of public debt to GDP must yield the total debt-to-GDP ratio for a given year:

Appendix 6.1 provides a full description of how we computed and decomposed public debt to GDP in the simulations.

This decomposition allows a more realistic evaluation of the sensitivity of debt indicators to interest rate and exchange rate shocks than simply maintaining the assumption of fixed shares, which implies that debt is continuously rebalanced by recurrent debt repurchase and reissuance. The interest rate specification used in Equation (6.1) is the implicit or average rate for the debt category, and not the marginal rate. These are approximated by taking the moving average of the marginal interest rate in year t and the previous M–1 years, where M is the average maturity of the debt category measured in years.

Model and Estimation

The inputs to produce debt simulations according to Equations (6.1) and (6.2) are obtained from stochastic simulations that capture the volatility, correlation, and persistence of the required inputs and other variables considered important to determine the state of the business cycle.

Simulations are produced by fitting historical data to a vector autoregression (VAR) model, and then using the coefficients to produce a large number of stochastic projections. We specify the VAR as follows:

where Xt=(gt,itd,ηt,mt,τt,st)T is a (6 × 1) vector of endogenous variables containing real output growth, a short-term market-determined domestic interest rate, the change in the bilateral real exchange rate (versus the U.S. dollar), the change in real net capital inflows, the percent change in total revenues, and the percent change in primary expenditures, respectively. it* is the U.S. federal funds rate; A(L) is a polynomial in the lag operator of degree ranging from 1 to 4, depending on the country estimation output. B is a vector of coefficients; and εt is a vector of well-behaved error terms: εt ~ N(0, Ω). Employing growth rates has the advantage of dealing with stationary variables and avoiding any explicit assumption on trends.

VAR models for stochastic simulations of debt-to-GDP dynamics have already been adopted in the literature.11 The main differences in our specification relative to previous work are (1) the explicit introduction of financial variables—such as net capital inflows and the foreign interest rate—that play a substantial role in explaining the volatility in Caribbean countries; and (2) the distinct treatment of total revenues and primary expenditures. In fact, in the stochastic simulations, the primary balance is computed as a second step starting from the simulated growth rates of total revenues and primary expenditures. Keeping revenues and expenditures separated allows us to simulate the feedback effects of expenditure-smoothing policies.12

Monte Carlo Simulation

We use the fitted VAR model to simulate the effect of expenditure-smoothing fiscal policies (active). For comparison, we also simulate primary expenditures using the VAR-fitted equation for primary expenditures (passive). The passive simulations provide an agnostic diagnosis of the prevailing public debt sustainability conditions.

The simulations are computed by generating a sequence of random vectors ε^t+1,...ε^T such that ω[t+1,T],ε^ω=Wυτ, where υω ~ N (0,1) and W is the Choleski factorization of Ω = W′W. At every forecast period, we draw 2,000 random vectors ε^ω, while the VAR produces joint dynamic responses of all elements in X.13 As noted in Garcia and Rigobon (2005), in the passive simulation, the method is not sensitive to the ordering of variables in the VAR, as stochastic simulation results are shaped by the variance-covariance matrix of reduced-form errors Ω, which is unique. However, the active simulations are sensitive to ordering. By setting primary expenditures last in order, we minimize the contemporaneous (within-year) feedback from primary expenditures to the rest of the economy.

Computing the simulations for the fiscal policy framework proposed in this chapter requires two inputs. The first is the starting point of the primary expenditure path. The resulting level of primary expenditure represents the primary expenditure envelope for the next budget cycle. Second, a specification of the primary expenditure pattern in the future is necessary to calculate public debt projections over the medium term. For this, we use the assumption that primary expenditure remains constant in real per capita terms (see Equation (6.4) below). This exercise is performed for each of the 2,000 random draws using the debt accumulation identity, which results in a distribution of public debt outcomes at each projected horizon. The starting level of primary expenditure is then pinned down iteratively by backward induction, after the desired debt upper threshold, horizon length, and probability have been chosen. In this way, the next budget’s primary expenditure envelope is consistent with a debt-to-GDP upper threshold d¯ to be met T years ahead with probability p, where d¯, T, and p are parameter choices of the policymaker. We denote s˜t+1 as the primary expenditure level that is consistent with dt+Td¯ with probability p.

Notice that setting p = 0.5 is, by construction, the level of primary expenditure that stabilizes public debt to GDP in expected terms over the medium term. A higher p implies that the policymaker is seeking additional primary expenditure consolidation and, as a result, is choosing to set public debt to GDP on a downward trend in expected terms. A higher probability is therefore a more prudent choice, as it ensures that debt will remain within low bounds with a higher probability.

The choice of the medium-term horizon length, T, is also important. In this case, for a given debt threshold and probability choices, a longer horizon (high T) increases the smoothness of the primary expenditure path, as the need for adjustment is averaged away over more years. Too long a horizon, however, could undermine the credibility of the framework.

We propose a specification for the forward-looking primary-expenditure-smoothing path according to the following:

where g¯ is a “trend” growth rate of real GDP, which we set equal to the average population growth rate; gω is the simulated growth rate of real GDP; κ is a parameter representing the responsiveness of the primary expenditure growth rate to deviations of real GDP growth from trend growth; s˜t+1 represents the consolidation (if s˜t+1<0) of primary expenditures at time t + 1; while s˜ω=g¯,ω[t+2,T].

Setting κ= 0 corresponds to running no discretionary expenditure impulse: it implies that primary expenditure is projected to remain constant in real per capita terms after the initial consolidation occurring in year t+1. In other words, for κ = 0, the primary expenditure level at t+1 is set so that, if primary expenditure remains constant in real per capita terms for T years, public debt will remain below the chosen upper threshold d¯ with probability p in year t+T. A positive κ represents a countercyclical fiscal policy impulse, as it would result in additional primary expenditure growth opposite to that of real per capita GDP growth. Analogously, we take a negative κ to represent a procyclical fiscal policy impulse.

Results

Table 6.1 shows the results of our simulation exercises for a sample of Caribbean economies. For illustrative purposes, we set the proposal to a medium-term horizon of five years. Maximum debt threshold announcements for 2015 are purely illustrative. In the last three columns, we show the primary expenditure consolidation in the 2011 budget in percentage points of GDP that would be necessary to maintain the debt-to-GDP ratio below the illustrative threshold with probabilities of 50 percent and 70 percent.

Table 6.1Fiscal Framework Proposal Simulation for Caribbean Countries(Percent)
Estimated Debt/GDP in 2010Estimated Primary Balance/ GDP in 2010Expected Debt/GDP in 2015 under “Passive” BehaviorIllustrative Debt/GDP Threshold in 2015Probability Debt/GDP < ThresholdRequired Consolidation of Primary Expenditures in 2011
Acyclical (k = 0)Countercyclical (k = 0.5)Procyclical (k = −0.5)
(Percentage points of GDP)
Antigua and119−1.81601200.58.39.17.6
Barbuda0.710.311.19.4
Bahamas65−3.979600.50.81.10.4
0.71.92.41.3
Barbados122−1.21371300.57.87.67.9
0.710.310.410.2
Grenada1160.31351100.51.30.71.4
0.73.73.23.7
Jamaica1397.61171000.5−4.2−4.0−4.4
0.7−4.2−4.0−4.4
St. Kitts and184−1.92051800.511.111.710.5
0.713.614.313.1
St. Lucia80−3.287700.50.71.30.1
0.72.53.41.7
St. Vincent91−1.4115900.56.15.86.4
and the0.77.77.57.9
Grenadines
Sources: Country authorities; and authors’ estimates and projections.
Sources: Country authorities; and authors’ estimates and projections.

Among the countries reported, Jamaica is the only one that does not show a particular need for further consolidation in 2011. This is probably the result of the consolidation program that the government has undertaken since 2009. In fact, under the passive behavior, debt to GDP is predicted to decline from 139 percent in 2010 to 117 percent in 2015. In 70 percent of the simulated cases, Jamaica meets the target of a debt-to-GDP ratio below 100 percent if it follows acyclical (primary expenditure constant in real per capita terms) or moderately procyclical or countercyclical rules even in the case of a small expenditure increase in 2011.

For the remaining countries, our simulations predict an upward trend for debt-to-GDP ratios in the next five years under the passive primary expenditure behavior. For the Bahamas, Grenada, and St. Lucia, a primary expenditure consolidation of about 2 percentage points of GDP to 3 percentage points of GDP in 2011—together with acyclical or moderately procyclical or countercyclical primary expenditure rules—would be enough to maintain debt to GDP in 2015 below the level predicted for 2010 in 70 percent of the simulated cases. For Antigua and Barbuda, Barbados, St. Kitts and Nevis, and St. Vincent and the Grenadines, the framework signals a significant need for consolidation in the 2011 budget, in excess of 7 percentage points of GDP.

Figure 6.2 provides a series of public debt fan charts that underlie the results explained above, based on the simulation exercise. In particular, we proceed to simulate the proposed policy for a sequence of periods. The parameter choices remain the same as in previous sections. To this end, we perform recursive simulations, mimicking what would happen in practice as the public debt targeting framework is applied over subsequent annual budget cycles. First, we calculate the primary expenditure envelope for the next budget cycle (2011) using the same parameter choices as in Table 6.1, while allowing the VAR to generate the remaining vector variables for 2011. With the simulations for 2011, we compute the implied end-2011 public debt stock using the debt accumulation identity (Equation (6.1)). Notice that as each of the 2,000 simulations in the Monte Carlo experiment is subject to a random shock, the deficits and debt stocks for the period will vary across simulations, despite the use of the same framework for all of them. Then, for each simulation, we recalculate the primary expenditure for the 2012 budget period consistent with the same proposed framework parameters, allow the VAR to simulate the remaining variables, and calculate the public debt stock as of end-2012. We proceed in this way until the 2015 budget period.

Figure 6.2Fan Charts of Public-Debt-to-GDP Ratios under “Passive” and Public Debt Targeting

Source: IMF staff calculations.

Note: Variations in shading correspond to 50, 75, 90, and 95 percent confidence intervals.

Public Debt Targeting in the Business Cycle: A Quantitative Assessment

For reference, we compare the results outlined in the previous section with the following two alternatives: (1) primary expenditure trajectory under the historical behavior, as obtained from the unrestricted VAR, and (2) for the structural primary balance rule that targets the same public-debt-to-GDP ratio for comparability,14 with elasticities of real levels of revenues and primary expenditures to the output gap of 1 and 0, respectively.15 GDP gaps are estimated using the Hodrick-Prescott filter trend as a measure of potential GDP. We obtain quantitative measures for three different sets of results considered of interest: (1) the degree of primary expenditure smoothing; (2) the cyclical properties of primary expenditure vis-à-vis GDP and revenues; and (3) the public debt-smoothing properties.

Primary Expenditure Smoothing

The simulations show that the primary expenditure smoothing properties of public debt targeting compare favorably relative to a structural balance rule for this sample of countries. As an indicator of variability over time, we compute the growth rate of simulated real primary expenditures over the simulation years (2011–15) and the standard deviation of these growth rates for each of the 2,000 simulations. Then we calculate the coefficient of variation (CV) as the standard deviation scaled by the average growth rate over 2011–15, and average it over the 2,000 simulations. The results are presented in Table 6.2. The coefficient of variation of the growth rate of primary expenditures under public debt targeting is in general lower than under a structural balance rule.16 Under this particular parameter choice, only for two out of eight country cases, the CV under the public debt targeting framework proposal is higher than under a comparable structural balance rule, but not significantly so.17

Table 6.2Smoothing Properties of Public Debt Targeting

(Coefficient of variation of simulated real growth rates of primary expenditures)1

Historical Behavior2Structural Balance Rule3Public Debt Targeting4
Antigua and Barbuda25.717.63.4
Bahamas1.51.32.3
Barbados3.31.31.9
Grenada91.95.82.7
Jamaica2.32.21.7
St. Kitts and Nevis1.53.02.7
St. Lucia3.52.21.5
St. Vincent and the6.12.41.8
Grenadines
Sources: Country authorities; and authors’ estimates and projections.

Ratio of the standard deviation to the mean of the average real growth rate of simulated primary expenditures. Mean and standard deviation computed from a Monte Carlo experiment based on the VAR simulation. The simulations for the structural balance rule and the policy framework proposal are based on the same parameters and public debt target levels. In absolute terms.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal. All parameters are set as in Table 6.1.

Sources: Country authorities; and authors’ estimates and projections.

Ratio of the standard deviation to the mean of the average real growth rate of simulated primary expenditures. Mean and standard deviation computed from a Monte Carlo experiment based on the VAR simulation. The simulations for the structural balance rule and the policy framework proposal are based on the same parameters and public debt target levels. In absolute terms.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal. All parameters are set as in Table 6.1.

Cyclically adjusted primary balances also show that the cycle-smoothing properties of the debt targeting proposal are comparable with those of a structural balance rule. Table 6.3 reports the CV of cyclically adjusted primary balances, both for the structural balance rule and public debt targeting. The CV is computed as the standard deviation of the cyclically adjusted primary balances across the 2,000 simulations, and then scaled by the year average for each year of the simulation period 2011–15. The results show that for five out of eight cases, the structural balance rule has more variability in cyclically adjusted primary balances under a structural balance rule, in principle signaling more shock-absorbing capacity compared with public debt targeting.

Table 6.3Coefficient of Variation of Cyclically Adjusted Primary Balances

(Monte Carlo-based simulations, 2011-15)1

20112012201320142015Average

2011–2015
Antigua and Barbuda
Structural balance rule20.651.610.560.270.340.69
Public debt targeting30.410.530.550.400.620.50
The Bahamas
Structural balance rule27.262.832.820.761.323.00
Public debt targeting31.321.530.790.591.201.09
Barbados
Structural balance rule20.860.944.370.640.721.51
Public debt targeting30.343.520.902.279.123.23
Grenada
Structural balance rule20.240.240.240.220.300.25
Public debt targeting30.350.200.210.300.390.29
Jamaica
Structural balance rule20.530.710.580.200.280.46
Public debt targeting30.180.100.320.190.370.23
St. Kitts and Nevis
Structural balance rule20.351.260.690.250.330.58
Public debt targeting30.820.911.021.266.522.10
St. Lucia
Structural balance rule20.563.222.736.8813.244.94
Public debt targeting30.410.940.380.741.800.85
St. Vincent and the Grenadines
Structural balance rule21.150.660.611.473.231.43
Public debt targeting30.330.340.670.602.310.85
Sources: Country authorities; and authors’ estimates and projections.

The cyclically adjusted primary balances are calculated as revenues scaled by the output gap, minus primary expenditures, and then the difference is divided by potential GDP. The potential GDP and the output gap calculations are based on the HP filter. The coefficients of variation are calculated as the standard deviation of the cyclically adjusted primary balances over the 2,000 simulated series, divided by the average.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

All parameters are set as in Table 6.1.

Sources: Country authorities; and authors’ estimates and projections.

The cyclically adjusted primary balances are calculated as revenues scaled by the output gap, minus primary expenditures, and then the difference is divided by potential GDP. The potential GDP and the output gap calculations are based on the HP filter. The coefficients of variation are calculated as the standard deviation of the cyclically adjusted primary balances over the 2,000 simulated series, divided by the average.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

All parameters are set as in Table 6.1.

However, notice that these calculations are based on the same simulations as in Table 6.2, which in general showed a more stable primary expenditure pattern for the public debt targeting framework. This means that, from consolidating the results in Tables 6.2 and 6.3, it is possible to conclude that (1) revenues are, in general, also more stable under public debt targeting (numerator of the CV effect), and/or (2) average (or trend) real growth rates of revenues, GDP, and primary expenditures are also higher under debt targeting (denominator of the CV effect), resulting in lower coefficients of variation. This means that the stabilization impact of accounting for the indirect effects under public debt targeting appears to more than offset the direct impact of a structural expenditure rule, which allows more primary balance variability to cushion shocks.

The simulation results on expenditure smoothing confirm the key conceptual difference between public debt targeting and structural balance rules. A structural balance rule does not take into account the indirect non-Keynesian feedback to the rest of the economy. In other words, it implicitly assumes that fiscal policy credibility is high, in the sense that sovereign spreads, interest rates, foreign exchange markets, and capital flows are not perturbed by deficits and debt accumulation. It therefore remains appropriate for countries with full credibility (not countries included in this chapter), or for cases in which credibility problems are resolved solely by the implementation of a structural balance rule consistent with debt sustainability.

Cyclical Properties of Primary Expenditures

Primary expenditures show, in general, a similar degree of co-movement with GDP as a structural balance rule for the set of parameter choices in this experiment. We illustrate this point by computing the correlations of the real growth rates of primary expenditures with the real growth rates of GDP and government revenues for each of the 2,000 simulations. The average correlation is then displayed in Table 6.4. The correlations with GDP of both the structural balance rule and the framework proposal are, as expected, significantly lower than under the historical behavior. Moreover, the correlations with GDP of public debt targeting are higher than under the structural balance rule for five out of eight country cases, but in three of the five cases the difference in correlation is small.18

Table 6.4Cyclical Properties of Primary Expenditures

(Correlation of growth rates of primary expenditures with GDP and revenues)1

Correlations of Real Growth Rates of

Primary Expenditures and GDP
Correlations of Real Growth Rates of Primary

Expenditures and Revenues
Historical

Behavior2
Structural

Balance Rule3
Public Debt

Targeting4
Historical Behavior2Structural

Balance Rule3
Public Debt

Targeting4
Antigua and Barbuda−0.010.120.280.090.870.70
Bahamas0.36−0.450.29−0.150.640.37
Barbados0.330.040.170.030.840.32
Grenada0.150.01−0.080.650.960.46
Jamaica0.080.080.440.300.940.56
St. Kitts and Nevis0.780.000.020.490.940.66
St. Lucia0.70−0.41−0.43−0.140.940.77
St. Vincent and the Grenadines0.25−0.05−0.160.520.940.32
Sources: Country authorities; and authors’ estimates and projections.

Correlation within each simulation’s real growth rates of primary expenditure with GDP and revenues, then averaged for the 2000 simulations.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal in this paper. All parameters are set as in Table 6.1.

Sources: Country authorities; and authors’ estimates and projections.

Correlation within each simulation’s real growth rates of primary expenditure with GDP and revenues, then averaged for the 2000 simulations.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal in this paper. All parameters are set as in Table 6.1.

The correlation between the real growth rates of primary expenditures and revenues under public debt targeting is high, however. The average correlation for the country sample is 52 percent for the parameter choice explained. This is not necessarily a limitation of the proposal, as the framework parameters can be set to increase the smoothness of spending (for example, by extending the medium-term horizon beyond five years). In fact, a positive correlation between revenues and expenditures implies that there is a need for a degree of procyclical adjustment, which prevents public debt from growing above the markets’ comfort zone. This is, therefore, a concrete feature of the framework proposal, as it forces the policymaker to acknowledge and to make a choice on the trade-off between some degree of procyclical expenditure consolidation (so as to avoid too large a deficit and debt accumulation in bad states of the cycle) and public debt volatility.19

Public Debt Smoothing

The points made above raise the issue of the volatility of public debt outcomes when the proposed framework is applied in practice, as an indicator of the change in the sovereign’s creditworthiness. To illustrate the quantitative implications in terms of debt smoothing, we compute the dispersion of debt outcomes out of the simulation exercise when the policy framework is applied over a sequence of periods. The results are summarized in Table 6.5. The first observation is that the volatility of the debt-to-GDP ratio by 2015 under the framework proposal appears to be less than one-half of that under historical behavior. In addition, the parameter choice also implies lower average volatility in debt outcomes than under a structural balance rule, with the coefficient of variation being on average about three-fourths of that for the structural balance rule.

Table 6.5Expected Public Debt Level and Dispersion

(Based on simulations for 2015)1

Historical Behavior2Structural Balance Rule3Public Debt Targeting4
ExpectedStd. Dev.CVExpectedStd. Dev.CVExpectedStd. Dev.CV
Antigua and Barbuda160.442.30.26120.918.00.15114.311.00.10
Bahamas79.412.50.1664.36.70.1060.95.30.09
Barbados136.815.90.12125.612.80.10125.310.80.09
Grenada134.637.90.28108.97.90.07111.36.40.06
Jamaica116.926.00.22139.111.30.08110.95.40.05
St. Kitts and Nevis205.319.90.10174.712.70.07170.719.70.12
St. Lucia87.121.10.2484.08.40.1079.85.70.07
St. Vincent and the Grenadines115.015.70.14105.26.50.0694.94.30.05
Sources: Country authorities; and authors’ estimates and projections.

Public debt and standard deviation of public debt are presented in percent of GDP. Calculated based on a Monte Carlo experiment with 2,000 simulations. CV is the coefficient of variation, calculated as the standard deviation of the debt-to-GDP ratio divided by the mean.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal in this chapter. All parameters are set as in Table 6.1.

Sources: Country authorities; and authors’ estimates and projections.

Public debt and standard deviation of public debt are presented in percent of GDP. Calculated based on a Monte Carlo experiment with 2,000 simulations. CV is the coefficient of variation, calculated as the standard deviation of the debt-to-GDP ratio divided by the mean.

Primary expenditure simulations based on the estimated VAR coefficients.

Primary expenditure simulations based on a structural primary balance with output gap elasticities of revenues and primary expenditures of 1 and 0, respectively.

Primary expenditure simulations based on the policy proposal in this chapter. All parameters are set as in Table 6.1.

Notice that there are two opposite forces at play behind this result. On the one hand, smoother expenditures can result in larger fiscal imbalances and, therefore, higher public debt volatility, for a given pattern of revenues (direct or Keynesian effect). On the other hand, if expenditure smoothing results in smoother macroeconomic outcomes, including not only GDP but also other critical variables such as real exchange rates and interest rates, the volatility of the debt-to-GDP ratio might actually be lower (indirect or non-Keynesian effect). The simulation results appear to indicate that, as per the parameter estimates in the VAR, the second effect is relatively stronger: the proposed framework generates both smoother primary expenditures and lower volatility of public debt (Figure 6.2).

The fan charts in Figure 6.2 illustrate the main policy announcement under public debt targeting, together with the primary expenditure budget envelope for the next budget cycle. They show the projected trajectory of public debt and its potential dispersion from economic shocks.

This analysis emphasizes the importance of including a measure of debt volatility under public debt targeting. It is possible to develop a similar framework without the use of probabilistic outcomes to assess public debt dynamics (including, if necessary, an additional primary balance surplus amount for prudential purposes). However, such a framework would not capture the impact of shocks on debt dynamics in an appropriate way. This is because, under public debt targeting, the degree of primary surplus reset required after every year’s observed shocks is a function not only of the parameter choice but also of the sensitivity of public debt dynamics to shocks, and also of the volatility, persistence, and correlation of the main variables affecting debt dynamics. Furthermore, the trade-offs in terms of debt structure play a critical role in determining the appropriate fiscal stance (for example, all else being equal, requiring more expenditure consolidation if a large share of public debt is denominated in foreign currency). This implies that the economic consideration that is at the heart of the proposal, namely the trade-off between expenditure smoothing and debt smoothing, would not be accounted for appropriately without some measure of dispersion in public debt outcomes.

Limitations

The main limitation of the public debt targeting proposal is that it requires a sufficiently strong political and societal commitment to fiscal prudence. This limitation extends to fiscal rules and numerical constraints as well. Without a strong willingness and capacity to contain special interests, distributional conflicts, or other political economy pressures, the mere adoption of the framework cannot be expected to deliver the intended results. In other words, we take the view that the causality goes from societal commitment to the adoption of a fiscal framework or rule—either the one proposed here or any other—and not the other way around.

Moreover, even if such societal agreement exists, the framework application requires some more technical preconditions. Fiscal institutions need to be sufficiently developed to appropriately monitor and control the budget (budget processes, fiscal accounting and auditing practices, public financial management, and so on). In particular, it is critical that public debt creation outside the government level used in the framework formulation be addressed for the policy to be credible. For example, the observed changes in central government debt can be significantly larger than accounted for by the reported central government deficit, change in government deposits, prefinancing operations, and valuation changes. This suggests that other sources of debt creation, typically from public entities and other quasi-fiscal bodies, can be important. Ideally, and unlike the examples in this chapter, the framework should be applied to the consolidated government finances, or at least to the most aggregated government definition possible. The budget process and institutions would need to be effective limiting factors to the universe of public entities and bodies, particularly when they can be sources of public debt and contingent liabilities.

In addition, budgets would need to include sufficient provisions for unforeseen spending. For example, in the case of Caribbean countries, natural disasters—such as hurricanes and flooding—usually damage public infrastructure and private property, and typically require government expenditures in excess of the budget. For the framework to work in such an environment, budgeted primary expenditures should include allocations for such contingencies. Otherwise, the framework’s objective of enhancing credibility would be weakened, as would result from the need to compensate for additional spending needs with lower current or capital spending in other areas. Expanding the proposal with a specific countercyclical component in case of natural shocks could be a possibility in this regard.

The technical complexity of public debt targeting could be considered a limitation. It could be argued that market participants would need to be sufficiently sophisticated to affect the formation of expectations sought in the proposal. However, we do not believe that this is a major limitation. Indeed, the two main announcements are clear, easy to understand, objective, and practical operationally: a debt trajectory within narrow probabilistic bands and next years’ primary expenditure budget envelope. The four intermediate technical parameters, which play the role of mapping the two main announcements, are the complicated part. The need is therefore not necessarily for a framework that allows every taxpayer or bondholder to understand the mechanics of the framework, but for independent and technically able institutions (private and public) that can assess the consistency of the announcement and inform the public.

This is why an independent institutional body (such as a “Fiscal Council”) could be an important complement to a debt targeting framework. It could assess consistency and keep the public informed. External institutions could play a role in assisting with the assessment when deemed necessary. In addition, institutional arrangements would need to be set up so that the framework is stable over time, in the sense that changes in the key framework parameters—such as the horizon, the debt threshold level, and probability—do not change from one budget cycle to the next arbitrarily.

Discussion

Public debt targeting can be used as a transitional framework toward the adoption of a structural balance rule. Implicitly, a structural balance rule assumes that there are no significant sovereign debt or credibility problems that can trigger the non-Keynesian effects: expenditures are determined in relation to potential GDP and structural government revenues (consistent with public debt sustainability targets over the long term). The cyclical movements in revenues are not considered critical, and any resulting cyclical deficits can be covered with debt issuance. Implicitly, this assumes that a government is fully credible, in the sense that it can commit not to default on public debt regardless of the circumstances, and therefore faces an infinitely elastic financing supply at low risk premiums regardless of the fiscal deficit and the stock of public debt. As this is not a realistic assumption for many countries, a fiscal policy framework that seeks to smooth the business cycle needs to incorporate the indirect effects from deterioration of the sovereign’s creditworthiness in bad states of the cycle.

If public debt is high, access to financing may be too costly at the time it is needed the most (or just unavailable), and could have destabilizing indirect effects on the rest of the economy through increases in sovereign interest premiums and on credit and foreign exchange markets more generally. In the public debt targeting framework, these indirect effects are accounted for explicitly, and the choice of technical intermediate parameters allows the best compromise between avoiding sharp changes in government expenditure and avoiding sharp changes in public debt. As public debt targeting is consistently applied over time and public debt declines to sufficiently low levels, indirect effects would eventually become less important and countries could adopt structural balance rules.

The proposal also is consistent with political economy constraints on fiscal policy decisions. First, although the proposal finds its rationale in the constraints imposed by financial markets, it does not imply that the impact of political economy factors on fiscal performance is minimized in any way. On the contrary, the proposal takes as a fact the possibility of the existence of political economy factors as a reason for fiscal sustainability problems and excessive debt accumulation. If so, financial markets may end up setting the limit eventually. And second, a sufficiently high level of political and social consensus on the value of a sustainable fiscal stance is a core precondition if this proposal is expected to be effective. Moreover, even if social consensus is already sufficiently high, institutional (formal or informal) incentives might distort decisions away from the desired collective objective.

Conclusion

Avoiding the procyclical fiscal policy bias observed in most of the developing world has a direct impact on smoothing aggregate demand (Keynesian effects). However, eliminating altogether the procyclical fiscal policy bias may not always be optimal in terms of smoothing aggregate demand. This is because countries with sufficiently high public debt (those with marginal increases in public debt that have a first-order impact ON investors’ perception of sovereign risk) might find it optimal to pursue some degree of fiscal consolidation during recessions to minimize the resulting deterioration in creditworthiness. This can result in an increase in sovereign spreads and domestic interest rates, and all the associated collateral damage, including crowding out of private investment and consumption (non-Keynesian effects). If public debt is already sufficiently high, excessive deficit and debt accumulation in low states can drag an economy into a negative spiral of high interest rates, credit crunch, and output decline.

This chapter has proposed a fiscal policy framework that balances out these two opposite effects, based on setting primary expenditure budget envelopes consistent with probabilistic debt sustainability outcomes. The proposal is based on policymakers’ announcement of two main indicators: a medium-term debt trajectory with the corresponding probability bands and a budget envelope for primary expenditures in the next budget cycle. These two announcements are easy to communicate and simple to understand for the general public, which is critical in order to anchor expectations on a sovereign’s fiscally sustainable path (implied by a declining debt trajectory that is resilient to shocks) while also avoiding the need for sharp expenditure consolidation during bad periods of the cycle (which could also undermine fiscal sustainability prospects given possible social and political resistance). Overall, the framework proposal puts a premium on the need to assure financial markets of fiscal sustainability as key to smoothing the business cycle by taking into account the indirect (non-Keynesian) effects from deterioration in a government’s creditworthiness in lieu of focusing only on the direct impact of fiscal policy on aggregate demand.

Monte Carlo simulations for a sample of highly-indebted Caribbean economies show that, for a set of plausible parameter choices, the public debt targeting framework can reduce primary expenditure’s procyclical bias while reducing the volatility of public debt. Empirically, it is shown that for the Caribbean countries in the sample and within a VAR methodology, these indirect effects more than offset the direct impact of procyclical adjustments required to avoid excessive debt accumulation on aggregate demand.

The proposal has some operational advantages. First, possible transparency issues related to distinguishing permanent from transitory shocks (for example, as in structural balance rules) are eliminated. Instead, spending levels are determined based on projected public debt dynamics, which is simpler and more transparent, as these are anchored around the debt accumulation identity. Second, the methodology forces the use of a broad set of variables to determine appropriate primary spending levels that capture the state of the business cycle, given the wide set of indicators needed to produce debt projections (including interest rates and real exchange rates). Third, it tackles the impact of debt sustainability concerns on the business cycle directly, given the explicit commitment to public debt thresholds and a debt trajectory band. And fourth, the framework can be designed to allow for a cyclical space for discretionary primary spending, possibly including a state-contingent additional primary expenditure component, but without necessarily allowing full discretion.

Appendix 6.1. Sovereign Debt Dynamics

Debt-to-GDP dynamics takes into account that the government issues both bonds denominated in local currency and bonds denominated in foreign currency, as well as borrows from bilateral and/or multilateral lenders (which we call, for simplicity, “official debt”). Thus, the stock of debt at the end of period t, denoted by Dt equals the sum of the debt denominated in domestic currency, Dtd; the debt denominated in foreign currency (typically U.S. dollars), Dtf, converted to domestic currency at the existing nominal exchange rate, et (units of domestic currency per unit of foreign currency); and the official debt (also denominated in U.S. dollars) Dto:Dt=Dtd+etDtf+etDto. This relationship can be converted in terms of ratios to GDP by dividing both sides by nominal GDP (expressed in local currency), Yt:

where dtd=DtdYt,dtf=etDtfYt,dto=etDtoYt.

The dynamics of the domestic-currency-denominated debt is described by the government’s flow budget identity Dtd=(1+it)Dt1dαtβt, where it is the interest rate on local-currency-denominated government debt whose maturity is at time t, Bt is the primary balance, and αt ∈ [0,1] is the fraction of primary balance used to service the domestic debt exposure. The primary balance is defined as Bt = Tt, – Gt where Tt is the revenue collected and Gt represents the non-interest public spending. The budget identity can be expressed in terms of ratios to GDP by dividing both sides by Yt:

where yt is the growth rate of nominal GDP and bt is the ratio of primary balance to GDP. Deflating the nominal growth rate by the inflation rate, πt the flow of budget identity can be expressed in the terms below, for real growth rate, gt:

Analogously, the dynamics of the foreign-currency-denominated debt is described by the following first-order difference equation:

where itf is the nominal interest rate paid on government bonds denominated in foreign currency whose maturity is at time t, and βt ∈ [0,1] is the fraction of primary balance used to service the foreign commercial debt exposure. Multiplying both sides by et and dividing through by Yt yields the following:

where t represents the nominal exchange rate depreciation. By deflating both the nominal growth factor and the nominal exchange rate factor by the domestic inflation rate, the above expression can be expressed in terms of the real growth rate gt, the real exchange rate depreciation ηt, and the real interest rate paid on foreign commercial debt (1+itf)/(1+πt*):

The dynamics of the official debt-to-GDP ratio is analogous to the one above. A substantial difference is that typically the interest rate applied by bilateral or multilateral lenders, ito, is lower than the market rate:

Another difference across various components of sovereign debt is that the average maturity of each class of debt differs. We take these differences into account when we compute the interest rate paid in every period. Let md, mf, and mo be the average maturities of domestic, foreign, and official debt, respectively, and ιtd,ιtf, and ιto be the one-period government bond rate for the afore-mentioned classes of sovereign debt. Then:

The nominal interest rate paid on one-period commercial bonds denominated in foreign currency, ιtf, can be computed as the sum of an international risk-free rate, it*, and a spread σtf representing the country’s sovereign risk: ιtf=it*+σtf. Similarly, ιto=it*+σto, where typically the spread paid to multilateral and bilateral lenders, σto, is lower than that paid to international financial markets, σtf.

Finally, as information on the fraction of primary balance servicing each component of debt is not available, we assume that at time t, the fraction of primary balance that services a certain component of debt is proportional to the existing share of debt of a particular type over the total:

Summing up Equations (A2) to (A4) yields the initial Equation (A1).

References

    AizenmanJ.M.Gavin and R.Haussman1996Optimal Tax Policy with Endogenous Borrowing Constraint,NBER Working Paper No. 5338 (Cambridge, Massachusetts: National Bureau of Economic Research).

    AlesinaA. and R.Perotti1994The Political Economy of Budget Deficits,NBER Working Paper No. 4637 (Cambridge, Massachusetts: National Bureau of Economic Research).

    AtkesonA.1991International Lending with Moral Hazard and Risk of Repudiation,Econometrica Vol. 59 (4) pp. 106989.

    BackusD.P.Kehoe and F.Kydland1992International Real Business Cycles,Journal of Political Economy No. 100 pp. 74575.

    BalassoneF. and M.Kumar2007Cyclicality of Fiscal Policy,” in Promoting Fiscal Discipline Chapter 3 (Washington: International Monetary Fund).

    BergA. and J.Sachs1988The Debt Crisis, Structural Explanations of Country Performance,Journal of Development Economics No. 29 pp. 271306.

    BlanchardO.1985Debt, Deficits, and Finite Horizons,Journal of Political Economy No. 93 (April) pp. 22347.

    BornhorstF.G.DobrescuA.FedelinoJ.Gottschalk and T.Nakata2011When and How to Adjust Beyond the Business Cycle? A Guide to Structural Fiscal Balances,IMF Technical Note and Manuals 11/02 (Washington: International Monetary Fund).

    BulowJ. and K.Rogoff1989aSovereign Debt: Is to Forgive to Forget?American Economic Review No. 79 pp. 4350.

    CataoL. and B.Sutton2002Sovereign Defaults: The Role of Volatility,IMF Working Paper 02/149 (Washington: International Monetary Fund).

    CelasunO.X.Debrun and J.Ostry2007Primary Surplus Behavior and Risks to Fiscal Sustainability in Emerging Market Countries: A ‘Fan-Chart’ Approach,IMF Staff Papers Vol. 53 No. 3.

    ChariV. and H.Cole1995A Contribution to the Theory of Pork Barrel Spending,Federal Reserve Bank of Minneapolis.

    ChuhanP.S.Claessens and N.Mamingi1998Equity and Bond Flows to Latin America and Asia: The Role of Global and Country Factors,Journal of Development Economics No. 55 (April) pp. 43963.

    ClineW. and K.Barnes1997Spreads and Risk in Emerging Markets Lending,Institute of International Finance Institute of International Finance Research Paper No. 97–1 (Washington).

    EdwardsS.1986The Pricing of Bonds and Bank Loans in International Markets. An Empirical Analysis of Developing Countries’ Foreign Borrowing,European Economic Review No. 30 pp. 56589.

    EichengreenB. and A.Mody1998What Explains Changing Spreads on Emerging Market Debt: Fundamentals or Market Sentiment?NBER Working Paper No. 6408 (Cambridge, Massachusetts: National Bureau of Economic Research).

    GarciaM. and R.Rigobon2005A Risk Management Approach to Emerging Market’s Sovereign Debt Sustainability with an Application to Brazilian Data,NBER Working Paper No. 10336 (Cambridge, Massachusetts: National Bureau of Economic Research).

    GavinM.R.HaussmanR.Perotti and E.Talvi1996Managing Fiscal Policy in Latin America,Working Paper No. 325 Office of the Chief Economist (Washington: Inter-American Development Bank).

    GavinM. and R.Perotti1997Fiscal Policy in Latin America,” in NBER Macroeconomics Annual 1997 (Cambridge, Massachusetts: MIT Press).

    International Monetary Fund2010World Economic Outlook: Recovery Risk and Rebalancing World Economic and Financial Surveys (WashingtonOctober).

    KaminskyG. and C.Reinhart2000On Crises, Contagion and Confusion,Journal of International Economics Vol. 51 (June) pp. 14568.

    KaminskyG. and C.Reinhart and C.Vegh2004When It Rains It Pours: Procyclical Capital Flows and Macroeconomic Policies,NBER Working Paper No. 10780 (Cambridge, Massachusetts: National Bureau of Economic Research).

    KopitsG. and S.Symansky1998Fiscal Policy Rules IMF Occasional Paper No. 162 (Washington: International Monetary Fund).

    MassonP.1985The Sustainability of Fiscal Deficits,IMF Staff Papers Vol. 32 (August) pp. 577605 (Washington: International Monetary Fund).

    MinH.1998Determinants of Emerging Market Bond Spread: Do Economic Fundamentals Matter?Policy Research Paper No. 1899 (Washington: World Bank).

    PenalverA. and G.Thwaites2006Fiscal Rules for Debt Sustainability in Emerging Markets: The Impact of Volatility and Default Risk,Bank of England Working Paper No. 307 (London). http://papers.ssrn.com/so13/papers.cfm?abstract_id=933299.

    PerottiR.1997bFiscal Policy and Saving in Good Times and Bad Times,” in Promoting Savings in Latin America ed. by RicardoHaussman and HelmutReisen.

    ReinhartC.K.Rogoff and M.Savastano2002Debt Intolerance,unpublished (Washington: International Monetary Fund).

    TalviE. and C.Vegh2000Tax Base Variability and Procyclical Fiscal Policy,NBER Working Paper No. 7499 (Cambridge, Massachusetts: National Bureau of Economic Research).

    TannerE.2003Fiscal Rules and Countercyclical Policy, Frank Ramsey Meets Gramm-Rudman-Hollings,IMF Working Paper 03/220 (Washington: International Monetary Fund).

    TannerE. and I.Samake2008Probabilistic Sustainability of Public Debt: A Vector Autoregression Approach for Brazil, Mexico, and Turkey,IMF Staff Papers Vol. 55 No. 1.

Note that as the framework is set in terms of primary spending, automatic stabilizers on revenues and interest expenditures are allowed to play in full.

The possibility of a countercyclical impulse would find its rationale in Keynesian arguments for aggregate demand smoothing, such as from nominal rigidities. However, a procyclical discretionary impulse could in some cases be appropriate if it helps to contain the non-Keynesian effects of public debt accumulation.

However, if rules are more generally defined as institutional mechanisms aimed at supporting fiscal credibility and discipline, then the proposed framework could be considered a rule.

Balassone and Kumar (2007) show, for a sample of developed and developing economies, that the overall balance behavior over the business cycle is asymmetric, as these tend to be relatively more negative in bad times than they are positive in good times.

For example, mechanisms can be introduced for an ex post compensation for miscalculations about the state of the cycle, as, for example, in Switzerland. Also, a broader set of variables could be used to determine structural versus cyclical components of revenues or expenditures in addition to GDP. Discretionary expenditures could also be included, provided this introduction is attached to an over-the-cycle compensatory mechanism. See Bornhorst and others (2011).

This is not referring to the technical aspects and multiple methodologies available to identify cycle and trend. It refers to available methodologies that are weak in identifying the trend position at endsample points, which results in large practical miscalculations that can undermine credibility, even when the methodology remains unchanged. See Bornhorst and others (2011).

For example, the IMF’s 1998 World Economic Outlook shows that, for a sample of Asian countries, exchange rates and commodity prices are more important in assessing the cyclical components of revenues and expenditures than GDP fluctuations.

At the other extreme of the spectrum, simple debt rules may not be credible either, as they might be difficult to comply with under tail risks. For example, a simple commitment to reduce public debt within a specified deterministic path (for example, a debt ceiling) might prove politically and socially difficult as it implies a commitment to adjust negative shocks in full within the current year, forcing a strongly procyclical fiscal stance. This difficulty can undermine their credibility.

Individual-country forecast variance decompositions are derived from a two-lag six-variable vector autoregression featuring annual observations of the output growth rate, a short-term interest rate, the growth rate of the real exchange rate, the change in net capital inflows, the total revenue growth, and the primary expenditure growth, respectively.

We abstract from cross-foreign-currency-parity changes. This is not a significant simplification, as most countries’ foreign currency debt used in the sample is denominated in U.S. dollars, if not exclusively so.

The simulation results need to be taken with caution. Any VAR approach to policy simulation is inherently subject to the Lucas critique. In fact, when policies change, the behavior of market participants might change as well, possibly affecting the estimated coefficients. This limitation can be addressed using a general equilibrium model.

For the purposes of the simulations, we also use assumptions on the U.S. federal funds rate, the London interbank offered rate on the U.S. dollar, and U.S. inflation in line with the World Economic Outlook (IMF, 2010), and we fit the domestic consumer price index to a simple pass-through equation to project the domestic inflation rate.

Also, the simulations under the structural balance rule are based on the same long-term GDP growth and real interest rate assumptions.

The primary expenditure under a structural balance rule is calculated as structural revenues minus the primary surplus that targets the same public-debt-to-GDP target as in the framework proposal, for comparability. Structural revenues are calculated as the simulated revenues times the output gap (computed as the ratio of the GDP trend using the Hodrick-Prescott filter with λ = 100 divided by the GDP level).

The calculations of the CV reported in Table 6.3 exclude the growth rate of primary expenditures in 2011 because for both fiscal policy choices, it captures the initial consolidation required to set public debt on a sustainable path.

Notice that this is not a horse race between the two methodologies, as the results are conditional on the set of parameter choices for each.

These correlations include the initial adjustment for the 2011 budgets reported in Table 2.

Notice that the simulation experiment also shows that the average correlation between revenue and GDP growth rates for this sample of countries is 88 percent under a structural balance rule, which is very high. However, this is only the result of a simplistic identification strategy of cyclical revenues, based only on output gap measures with unit revenue elasticity as explained above.

    Other Resources Citing This Publication