Professional Documents
Culture Documents
Cataloging-in-Publication Data
Joint Bank-Fund Library
Names: Eyraud, Luc. | Baum, Anja, 1985- | Hodge, Andrew. | Jarmuzek, Mariusz. | Kim,
Young. | Mbaye, Samba. | Türe, Elif. | International Monetary Fund. Fiscal Affairs
Department.
Title: How to calibrate fiscal rules : a primer / this note was prepared by a team led by Luc
Eyraud and including Anja Baum, Andrew Hodge, Mariusz Jarmuzek, Young Kim, Samba
Mbaye, and Elif Ture.
Other titles: How to notes (International Monetary Fund)
Description: [Washington, DC] : Fiscal Affairs Department, International Monetary Fund,
2017. | How to notes / International Monetary Fund | December 2017. | Includes
bibliographical references.
DISCLAIMER: Fiscal Affairs Department (FAD) How-To Notes offer practical advice from
IMF staff members to policymakerson important fiscal issues. The views expressed in FAD
How-To Notes are those of the author(s) and do not necessarily represent the views of the
IMF, its Executive Board, or IMF management.
This note provides guidance on how to calibrate General Principles for Calibration of Rules
fiscal rules; that is, how to determine the thresholds The calibration methodology presented in this note
(ceiling, floor, or target) for specific fiscal aggregates is based on four general principles:1
constrained by rules. The note focuses, more specif- •• Calibration should be comprehensive and consis-
ically, on the calibration of the debt, balance, and tent. Most countries have more than one fiscal
expenditure rules. rule (Schaechter and others 2012). To minimize
It is one of two guidance notes on the design of fis- the risks of inconsistency and conflict among
cal rules; the other note focuses on rule selection (IMF, rules, the fiscal framework should be assessed as a
2018). The two exercises are linked: if a fiscal frame- whole and the thresholds should be calibrated in
work had to be built from scratch, rules would need to a consistent manner. In particular, there should
be selected and calibrated at the same time. However, be a clear relationship between the debt and fiscal
to simplify the analysis and because some countries balance ceilings.
already have fiscal rules embedded in their laws, this •• Calibration should be sequenced. By analogy with
note examines the issue of calibration on its own. monetary policy, a well-designed fiscal framework
The note is not exhaustive or definitive. There are could set targets for both final and intermediate
many approaches to determining thresholds for rules. objectives of fiscal policy. Therefore, the frame-
This note presents a selection of methods that are work could be structured around two types of rules
intuitive, are simple to implement, and leave room for (Eyraud and Wu 2015). The first type, called fiscal
policy judgment. The methods are based on past IMF anchor, ensures that the framework achieves its
work, including analytic and policy papers, technical final objective, which is to preserve fiscal sus-
assistance missions, and training. Future work might tainability; a natural candidate is a ceiling on the
enhance or modify the framework presented here. debt-to-GDP ratio.2 The framework should also
Another important point is that the objective of this include operational rules that are under the con-
note is to provide practical guidance; it is not a substi- trol of governments while also having a close and
tute for the use of full-fledged macroeconomic models, predictable link to debt dynamics (for instance, a
which have the potential to capture all macroeconomic ceiling on the fiscal deficit or a cap on expenditure).
and fiscal implications of alternative calibrations. One implication of this dual structure is that the
The note is divided into four sections. The first sec- calibration exercise should be sequenced (see Box 1
tion discusses general principles used to calibrate rules. for a summary of the overall approach developed in
The second section reports international evidence on this paper). The debt ceiling should preferably be set
the numerical ceilings used in existing rules. The third first, taking into account debt sustainability and the
and fourth sections provide guidance on the calibration need to protect the country against adverse shocks.
of the debt ceiling and the operational rules (fiscal bal- Then the operational rules can be calibrated from
ance and expenditure rules). EViews econometric and the debt ceiling.
Excel files and manuals accompany this note to assist
economists with country-specific calibration exercises.
They are available upon request from the authors.
1These principles apply to the exercise of rule calibration, not rule
Method 1. Method 2.
When maximum debt limit is known When maximum debt limit is unknown
Method 1. Method 2.
Hitting the debt target in the long term Hitting the debt target by a given date
Method 3. Method 4.
Hitting the debt target by a given date Hitting the debt target by a given date
following a transition period after a period of fiscal buffer buildup
Method 1. Method 2.
Expenditure growth rule Expenditure ratio rule
•• Calibration should be prudent. Governments should cally updated. But the update process should not be
take fiscal risks into account in setting fiscal targets too frequent; what separates rules from annual bud-
and should preserve buffers to accommodate get targets is that rules are long-lasting constraints
shocks. Not all risks can be mitigated, insured, on fiscal policy. According to the IMF definition,
or provisioned for through contingency funds in rules are numerical targets fixed in legislation that
the budget; therefore, it is essential to create fiscal are binding for a minimum of three years. A review
headroom by setting prudent debt and deficit ceil- of fiscal rules could, for instance, be part of the
ings (IMF 2016). electoral cycle (Schaechter and others 2012).
•• Calibration should be updated regularly but not too
frequently. Fiscal rules are designed to be robust to
macroeconomic shocks, but conditions may evolve International Experience on Fiscal
over time. For example, a government might change Rule Thresholds
the way it responds to debt developments and the Fiscal rules have become more common worldwide
output gap. Countries that have had procyclical over the past two decades. In the early 1990s, only five
fiscal policies in the past may gradually move toward countries had fiscal rules in place. By the end of 2015,
a countercyclical stance (Frankel, Vegh, and Vuletin the number of countries with at least one national or
2013). The calibration should therefore be periodi- supranational fiscal rule surged to 92, of which more
than 60 percent were emerging market and developing supranational rules, the 60 percent threshold remains
economies.3 the most common among national debt rules.
Ceilings on public debt are a common feature of Most fiscal frameworks rely on additional rules to
rule-based fiscal frameworks. As of 2015, about 70 operationalize their debt ceilings. This occurs because
countries worldwide had a fiscal framework with an debt trajectories are not directly controlled by poli-
explicit cap on public debt. Debt rules are generally cymakers. More than 80 percent of countries with a
set in gross rather than net terms (gross debt minus debt ceiling also have rules imposing constraints on the
financial assets) because it is hard to determine which (nominal or structural) budget balance; among those,
government assets are truly liquid, particularly in times almost a third also have expenditure ceilings in their
of financial stress. Also, the concept of net debt is less fiscal frameworks. For nominal budget balance rules,
transparent than gross debt and more difficult to com- the 3 percent deficit ceiling is dominant through-
municate to the public. out the world; it has been adopted by the European
Gross debt ceilings can vary significantly across Union, the West African Economic and Monetary
countries, but frequently range between 60 percent Union, and the East African Community (Figure 2).
and 70 percent of GDP (Figure 1). The clustering of With regard to structural budget balance rules,4
ceilings around these values reflects the strong repre- ceilings have a wider distribution (Figure 3). The con-
sentation of supranational rules. About three-quarters centration of deficit ceilings between 0 and 1 reflects
of countries with a debt ceiling are members of supra- the adoption of medium-term budgetary objectives
national unions. For instance, the European Union (MTOs) in European Union countries, as well as the
and the Eastern Caribbean Currency Union impose a structural balance ceiling (using average oil revenues
debt ceiling of 60 percent of GDP, while the Central instead of actual oil revenues) used in the Central
African Economic and Monetary Community and the African Economic and Monetary Community. A few
West African Economic and Monetary Union both countries impose two structural deficit ceilings—one
impose a cap of 70 percent of GDP on public debt. at the supranational level and another at the national
In the East African Community, member countries level—possibly with different thresholds. For instance,
have adopted a debt ceiling of 50 percent of GDP in Germany has structural deficit ceilings of 0.5 percent
net present value terms during the convergence process of GDP (supranational medium-term objective) and
toward the East African Monetary Union. Excluding 0.35 percent of GDP for the federal government
(national rule).
3The stylized facts presented in this section are based on the 2017
version of the IMF Fiscal Rules Database, available at http://www 4In this note, structural balance rules are defined as rules that
.imf.org/external/datamapper/fiscalrules/map/map.htm. It covers 96 correct the nominal fiscal balance for cyclical and transitory factors
advanced, emerging, and developing economies. (including, in some cases, the commodity price cycle).
12
10
Number of countries
0
–0.50 –0.25 0.00 0.45 0.50 1.00 1.25 1.70 2.00
Structural deficit (percent of potential GDP)
Source: IMF Fiscal Rules Database.
Expenditure rules are less common than debt or long-term nominal GDP growth (European Commis-
budget balance rules. These rules appear in several sion, 2016). One of its characteristics is that it also
forms (Cordes and others 2015). In a majority of cases, corrects for revenue measures, meaning that spending
expenditure rules consist of an explicit cap on nom- growth cannot exceed medium-term growth unless the
inal or real expenditure growth. Expenditure growth additional spending is matched by new discretionary
is capped either by a fixed numerical ceiling (in the revenue measures.
range of 2 percent to 4 percent for real growth rules in
our sample) or by a measure of medium- to long-term
growth. In a few emerging market economies, expen- Calibrating the Public Debt Ceiling
diture rules apply to the expenditure-to-GDP ratio, This section outlines how to set the ceiling of the
with ceilings in the range of 30 percent to 40 percent fiscal anchor (the debt ratio), with an approach based
of GDP. At the supranational level, only the European on precautionary considerations. Countries face sudden
Union has imposed an expenditure rule on member increases in debt because of negative macroeconomic
states. This rule, called the expenditure benchmark, shocks and the realization of contingent liabilities.
caps the annual growth of primary expenditure with One way to calibrate the debt ratio is to ensure with
high probability that debt is kept under control despite despite the potential for negative shocks. The debt rule
these shocks (for a general discussion, see IMF 2016). ceiling is computed as the maximum debt limit minus
Specifically, this section presents two alternative the safety margin.
methods. The first method assumes that there is This method was developed by Debrun and others
a known maximum debt limit beyond which debt (2017) and Baum and others (2017). Earlier research
dynamics spiral out of control. This method was on stochastic simulations of debt trajectories includes
developed by Debrun, Jarmuzek, and Shabunina IMF (2003a), Ferrucci and Penalver (2003), Garcia
(2017) for advanced and emerging market economies, and Rigobon (2004), and Celasun, Debrun, and Ostry
while Baum and others (2017) adapted the method to (2007). The following paragraphs describe the three
low-income countries. In this approach, the debt rule steps in greater detail.
ceiling should be set low enough to ensure with high
probability that debt will remain below the debt limit Step 1: Setting the Debt Limit
even when negative shocks occur. The second method In the first step, an assumption is made on the
does not rely on an explicit debt limit; it selects the maximum debt limit, which is country‑specific and
debt ceiling so that debt can be stabilized following can depend on many factors. There are many ways to
negative shocks without breaching a policy limit—a set the maximum level of debt that a country does not
maximum feasible level of the primary balance. wish to cross. Given the lack of consensus in the lit-
The debt ceiling set using either of these meth- erature, a sensitivity analysis based on alternative debt
ods can be based on any institutional coverage (for limit estimates could be warranted. The following are
example, central or general government).5 However, two possible approaches.6
it is important to ensure that the same institutional •• Risk of debt distress. The debt limit could be thought
coverage is used throughout the calibration exercise. of as the level beyond which it is believed that a
In other words, if the debt ceiling is computed for the debt distress episode will occur with heightened
central government, the related deficit and expenditure probability (for example, default, restructuring, or
ceilings will also apply to the central government. large increases in sovereign spreads). For instance,
This section presents the two methods in an intui- for emerging market economies and advanced
tive way. More details on the algorithms and formulas economies, the IMF Debt Sustainability Analysis
are provided in Appendix 1 and the manuals accompa- (DSA) framework for Market Access Countries uses
nying this note. benchmarks of 70 percent and 85 percent of GDP,
respectively (IMF 2013a). For low-income countries,
the IMF DSA framework has benchmarks for public
Method One: Calibrating the Debt Rule Ceiling When the debt in nominal terms in the range of 49 percent
Maximum Debt Limit Is Known to 75 percent,7 depending on institutional quality
In this method, stochastic simulations are used to (IMF and World Bank 2012).
calibrate the debt rule ceiling by computing a safety •• Risk of growth slowdown. The debt limit can be cho-
margin below a known debt limit. The calibration is sen as the level beyond which it is believed growth
done in three steps. The first step is to identify the will decline. For instance, Cecchetti, Mohanty, and
maximum debt limit. Second, the distribution of Zampolli (2011) find that debt becomes a drag on
macroeconomic and fiscal shocks is estimated and growth when it exceeds around 85 percent of GDP
used to simulate potential debt trajectories over a in Organisation for Economic Co-operation and
medium-term projection horizon. The results of these Development (OECD) countries. See Cottarelli
simulations are summarized in a fan chart. The third
step identifies the debt rule ceiling, which is a suf-
ficiently low starting level for debt (in the first year
6For a conceptual framework on debt limits as well as empirical
of the projection horizon) such that there is a safety estimates, see Ghosh and others (2013).
margin and debt will remain below the maximum debt 7The corresponding benchmarks for the present value of public
limit over the medium term with high probability, debt range from 38 percent to 74 percent of GDP: IMF(2013b).
For low-income countries, the present value of debt can differ from
its nominal value, particularly in countries that rely on concessional
5IMF (2009) provides guidance about the appropriate institu- external debt, where the nominal value may be a poor indicator of
tional coverage of fiscal rules. the debt service burden in the near term.
and Jaramillo (2012) for further discussion of the produces a path for macroeconomic variables over a
relationship between public debt and growth. medium-term projection horizon, where the variables
have been subject to shocks in each period.
Step 2: Estimating the Effect of Shocks on Debt Medium-term debt trajectories consistent with
The second step is to perform stochastic simulations each simulated path of macroeconomic variables are
to gauge the potential impact of macroeconomic and obtained from the system of simultaneous equations
fiscal shocks on debt over the medium term. This formed by the debt accumulation equation (that is, the
requires estimating the joint distribution of macro- government budget constraint) and a Fiscal Reaction
economic variables (Box 2).8 The set of variables to Function (FRF) in which the level of the primary bal-
include in the joint distribution varies by country ance may respond to the level of debt and realizations
group. For advanced and emerging market economies, of macroeconomic variables (see Box 3 for a discussion
the econometric files accompanying this note use GDP of the choice of the function). The fiscal reaction func-
growth, interest rates on government debt, and the tion includes a fiscal shock realized each period.10 Debt
exchange rate; for low-income developing countries, trajectories produced using stochastic simulations can
the terms of trade gap and disbursements of foreign be summarized in a fan chart.
loans are also included. Multiple simulations are car-
ried out using the joint distribution.9 Each simulation Step 3: Calibrating the Debt Ceiling
The final step is to calibrate the debt rule ceiling.
8The joint distribution can be specified using either a Vector This is done by choosing an initial level of debt (in the
Autoregression (VAR) or by directly calibrating a joint normal (or first projection year) so that debt remains below the
Student’s t) distribution based on historical co‑movement between maximum debt limit with a chosen probability over
variables. Alternatively, if these variables are expected to have differ-
the medium term, despite the occurrence of negative
ent means to those observed in the past, the programs attached to
this note allow the user to specify manually the mean of the forecast
variables around which shocks are added. See Box 2 and Appendix 1
for more details. level of debt. For example, an adverse shock on the interest-growth
9In both Methods One and Two, the macroeconomic shocks are differential will increase debt by more when the initial debt
drawn from symmetric (normal or Student’s t) distributions. This level is higher.
does not reflect empirical evidence that shocks can be skewed to the 10The distribution of fiscal shocks is calibrated based on deviations
downside in reality (Escolano and Gaspar 2016). Nonetheless, the between actual fiscal responses observed (that is, actual levels of
fan charts may be asymmetric, because the impact of the shocks on the primary balance) and the fiscal response predicted by the FRF
debt paths (in the form of automatic debt dynamics) depends on the over the sample.
( 1r 21 gg )
5 (1 2 1) ____
2 __
ai
d*
(3.3)
It is also possible for the user to impose a mean
value for the primary balance in each period over the
The long‑term steady state values for growth and medium-term horizon. This may correspond to the
the real interest rate can be proxied by imposing user’s baseline forecast (for instance, based on IMF
steady state on an estimated VAR model of the type World Economic Outlook projections). A fiscal shock
described in Box 2 (that is, imposing that lagged can be added to this mean value each period to cap-
values of variables must equal current values) and then ture uncertainty about future fiscal behavior.
50
Safety
45 margin
40
35
Debt rule ceiling
30
25
20
2016 17 18 19 20 21 22
Source: IMF Staff Calculations.
Note: Debt is projected over a six-year horizon from 2017-2022. The last year before projections begin is 2016.
shocks.11 Changing the starting debt level would shift the level of risk a government is willing to accept. A
and tilt the entire fan chart. The tolerated probability probability of 10 percent is used as a baseline. Figure 4
of breaching the debt limit can be chosen based on shows an example of a debt ceiling calibration over a
six-year horizon, with a safety margin below the max-
11The stochastic simulations are based on symmetric laws and
imum debt limit. If the current debt level is above the
draw both positive and negative shocks, but only negative shocks
matter for the calibration method. Indeed, the debt ceiling is cali- debt ceiling, then the country is not maintaining a suf-
brated so that debt remains below the maximum debt limit, except ficient safety margin given the degree of risk tolerance.
in a small percentage of cases when particularly bad combinations of The debt ceiling and the size of the safety margin will
shocks are realized.
be sensitive to a number of key parameters. For exam- Method Two: Calibrating the Debt Rule Ceiling When the
ple, the safety margin will be larger if: (1) the level Maximum Debt Limit Is Unknown
of risk aversion12 increases (as debt must be lower to The second method calibrates the debt ceiling in
reduce the probability of breaching the limit); (2) the cases in which the maximum debt limit is unknown.
amount of macroeconomic volatility implied by the This method is most suitable for advanced economies
estimated joint distribution increases (because larger with unconstrained market access, where considerable
shocks can generate larger increases in debt, a larger uncertainty might exist about how much debt can be
safety margin is required); (3) the response of the sustained.13
primary balance to changes in debt becomes weaker, At the core of this method is the assumption that
as reflected in the parameters of the FRF (a larger the primary balance is bounded upward. A primary
safety margin is required when the government is not surplus above a certain bound may be unachievable
acting strongly enough to offset the impact on debt of for various reasons, including political and public
negative shocks); or (4) the length of the medium-term resistance to spending cuts or the fact that additional
projection horizon increases (over a longer horizon, a revenue-raising measures eventually become ineffective
larger margin is required to ensure with high probabil- (the peak of the Laffer curve has been passed). Given
ity that debt remains below the limit, as there is greater that countries cannot promise to do whatever it takes
uncertainty about outcomes farther into the future). to ensure debt sustainability under all circumstances,
An important question is how to deal with contin- there is necessarily a level of public debt, above which
gent liabilities, which can be significant and relatively debt stabilization becomes “impossible.”
frequently realized (Bova and others 2016). In the The method is implemented in three steps. First,
methodology presented above, debt projections will the maximum feasible primary balance is identified.
likely reflect “normal” contingent liability realizations Second, stochastic simulations are used to obtain
through two channels: multiple trajectories of macroeconomic variables, the
•• Above-the-line contingent liabilities. The FRF primary balance, and debt, using an FRF that describes
includes the impact of fiscal shocks. These shocks, fiscal behavior in response to debt (see the options in
which are drawn from their historical distribution, Box 3). Finally, the debt rule ceiling is identified as the
are affected by the materialization of past contingent initial value of debt, to ensure with high probability
liabilities, provided they were recorded above the that debt can be stabilized following negative shocks
line (generally, under transfers) and transmitted to without breaching the maximum feasible primary
the primary balance. balance.14 The following paragraphs describe the three
•• Below-the-line contingent liabilities. The debt accu- steps in greater detail.
mulation equation includes stock-flow adjustments,
which are simulated over the forecast horizon using Step 1: Setting the Maximum Primary Balance
their historical distribution in the codes accompany- The first step is to identify the maximum fea-
ing this note. Thus, debt simulations will reflect the sible primary balance.15 The calibration can be
historical pattern of the realization of contingent lia-
bilities, provided they were recorded below the line.
13This method was developed by an IMF staff team that included
If the researcher expects contingent liabilities—
X. Debrun, M. Jarmuzek, C. Lonkeng, S. Basu, N. End, W. Shi, J.
either above or below the line—to be larger than those Sin, and F. Toscani.
experienced historically, this information could be 14Caution should be applied when using Method Two with an
captured by manually adjusting the stock-flow adjust- estimated (rather than normative) fiscal reaction function. If the esti-
mated reaction function is explosive (reflecting undisciplined fiscal
ments in the debt accumulation equation (see manuals behavior in the past), the initial debt level, computed with Method
attached to this note). Two, cannot be considered safe, given that it would place debt on an
unsustainable path. In this case, the normative or calibrated reaction
function should be used.
15In this method, it is possible to back out the debt level con-
sistent with the maximum feasible primary balance, but this debt
level is not an absolute upper limit for debt as in Method One. It is
12In this simple framework, a higher degree of risk aversion is the highest debt level that can be stabilized without breaching the
equivalent to a smaller required probability of debt breaching the maximum feasible primary balance given current macroeconomic
maximum limit following negative shocks (for instance, a threshold conditions. If the interest-growth differential increases, the maximum
of 5 percent or 1 percent rather than 10 percent). debt mechanically declines.
4 80
–4 40
–8 20
2016 17 18 19 20 21 22 2016 17 18 19 20 21 22
Source: IMF Staff Calculations
Note: The simulations are based on a normative fiscal reaction function. The primary balance and debt are projected over
a six-year horizon from 2017-2022. The last year before projections begin is 2016.
lower debt reduces the probability of exceeding the may be skewed in the adverse direction. Also, the
maximum primary balance. A lower debt ceiling will distribution of the shocks may not capture tail
also be implied by higher macroeconomic volatility, events well.
since the primary balance would need to be higher to •• Data constraints may prevent precise estimation
stabilize debt when negative shocks are larger. Finally, of the VAR or calibration of the joint distribution,
a lower long-term target level of debt may require a if only short time series for the relevant variables
lower debt ceiling, in cases where the normative fiscal are available.
reaction function is used (see Box 3). The normative •• The fiscal reaction function used to project fiscal
FRF will embody a stronger response of the primary variables contains only a small set of independent
balance to the current level of debt when the long‑term variables and ignores potential nonlinearities and
target is lower; so maintaining a lower debt ceiling breaks in the reaction of fiscal policy to debt.
will help keep the required primary balances below the •• Fiscal variables (for example, the primary balance
maximum feasible. or debt) are not included in the VAR or the joint
distribution, meaning that there is no feedback from
fiscal policy changes to macroeconomic variables, in
Limitations of the Proposed Approach particular GDP.
The tools proposed in this note do not substitute for
the use of full-fledged macroeconomic models that can Another important limitation is that the calibration
better capture structural and nonlinear relationships methods presented above rely exclusively on the need
among variables. This framework for calibrating the to protect a country’s fiscal position against negative
debt ceiling is made tractable by several simplifying shocks. Maintaining prudent debt levels is crucial to
assumptions:16 guarantee fiscal sustainability, but this is not the only
•• Macroeconomic variables are projected using a criterion that can or should be taken into account
simple VAR econometric model subject to shocks when setting a debt target. In many countries, increas-
(drawn from a symmetric joint normal/Student’s t ing public investment and funding education and
distribution) or by drawing directly from the joint health care are also priorities, and at least part of these
distribution if quarterly data are unavailable. Thus, expenditures must be financed through public debt.17
the macroeconomic simulations are informed by his- Therefore, there is an inherent trade-off when
torical data and cannot capture the impact of recent deciding on the level of the debt ceiling. Higher debt
or expected structural changes in the economy. increases vulnerability to shocks and can undermine
•• Structural/behavioral equations from economic market confidence and lead to fiscal distress. But the
theory (for example, aggregate demand curve or debt ceiling should not be too low, to allow space for
monetary policy rule) are not used to project mac- financing development needs. More complex models
roeconomic variables, potentially missing valuable are needed to reflect the trade-off between risk man-
information on how the economy would behave agement and development perspectives. By ignoring
in the future. this trade-off, the methods proposed in this note may
•• A VAR econometric model has a simple linear be biased toward austerity, at least for some emerging
structure. This structure may fail to capture non- and developing economies.
linearities among macroeconomic variables, such as
changes in the relationship between interest rates
and growth throughout the business cycle.
•• The VAR and the joint normal/Student’s t distri-
bution are based on the assumption that macro-
17Economic theory suggests that public investment should be
economic shocks are symmetric. In reality, shocks
primarily financed by debt issuance rather than taxes (Ostry, Ghosh,
and Espinoza 2015). One reason is that the distortions brought
16In the econometric files accompanying this note, users can make by taxation are smaller when tax increases are smoothed over time
ad hoc adjustments to forecasts of macroeconomic variables and through debt finance. In addition, public investment projects are
the primary balance to incorporate country-specific knowledge and expected to generate gains over several years and to benefit future
judgment, which may help overcome some of the limitations. See generations; therefore, their full cost should not be borne only by
Boxes 2 and 3 for more information. current taxpayers.
Framework for Commodity Exporters in resource wealth. These models formalize the
In the methods presented above, the fiscal rule asso- trade-off faced by commodity producers between
ciated with the final objective of fiscal policy (that is, investing resource revenues in real or financial assets.
debt sustainability) is the gross debt rule. But net debt A higher targeted stock of real assets is likely to
(debt minus financial assets) may be a more relevant come at the expense of a lower stock of net finan-
anchor in countries with large financial assets. For cial assets. Models with fiscal multipliers somehow
instance, in resource-rich countries it is common to mitigate this trade-off, because higher public invest-
focus on net wealth, measured as net financial wealth ment generates higher financial savings through its
(financial assets minus debt) plus resource wealth—the positive impact on GDP growth and nonresource
present value of future resource revenues (Baunsgaard revenues. In these models, a rule on net wealth can
and others, 2012). The calibration of the net wealth be introduced only after the scaling-up period of
(or net financial wealth) target can be based on one of investment; that is, once net wealth stabilizes again.
the two following approaches: Risk-Based Approach
Long-Term Sustainability Approach Uncertainty is another important consideration
An important challenge is to decide how to allo- when calibrating net wealth targets in resource-rich
cate net wealth across generations, given that natural countries. These countries need larger and more
resources are exhaustible. Various models have been durable buffers than other countries because economic
developed to calibrate the appropriate level of net shocks can be large and highly persistent. The amount
wealth in resource-rich countries, assuming the need to of the required savings depends on the degree of
achieve fiscal sustainability and intergenerational equity resource dependence, the level of risk the country is
(see a review in IMF 2012). facing, and its risk tolerance.
•• Fixed net wealth benchmark. In the standard model Several methods exist to compute the level of net
of the Permanent Income Hypothesis (PIH), financial wealth that countries should maintain as a
intergenerational equity is achieved by preserving precautionary buffer—a buffer that can be tapped in
government net wealth at its initial level, so that bad times to support spending when resource revenues
future generations will enjoy a similar amount of fall short. For instance, IMF (2012) proposes to use
wealth as the current generation. Several variations a value-at-risk (VaR) approach and a model-based
of this model exist, depending on whether net approach to estimate the adequacy of buffers. Both
wealth is preserved in real terms, real terms per methods estimate the minimum buffer that can absorb
capita, or as a share of GDP. The general idea is tail risks in resource revenue volatility. Specifically, the
that if only a fraction of resource revenues is spent buffer should be set high enough to ensure with high
every period, financial savings will increase suffi- probability that it is not fully depleted over the forecast
ciently to make up for the depletion of resource horizon and, therefore, expenditure cuts will not be
wealth. Total net wealth is therefore kept constant, needed. Another method was developed in “The Com-
although its composition changes over time: the modities Roller Coaster” (IMF 2015): the amount of
share of resource wealth (present value of future financial savings is calibrated to ensure that investment
resource revenues) will decline over time, but this returns on financial assets are sufficient to avoid large
decline will be perfectly offset by an increase in net fiscal adjustment in the event that commodity prices
financial wealth. fall. To illustrate, IMF (2015) computes, for three
•• Variable net wealth path to accommodate higher major oil exporters, the level of financial assets that
investment. Alternative models, also based on the would be sufficient to generate investment returns to
intertemporal budget constraint, relax the assump- cover half the lost revenue over five years with 75 per-
tion of constant net wealth. Extensions of the PIH cent to 90 percent probability.
model allow for a temporary scaling-up of public
investment. Net wealth initially declines during From the Debt Rule to the Operational Rules
the scaling-up period, because saving is lower than
under the PIH and the stock of financial assets does Debt and deficits are tied through an accounting
not increase quickly enough to offset the decline identity. A country’s debt is the cumulative stock of
past deficit flows, while the (overall) deficit captures
the annual change in the country’s debt. In practice, creating space for long-term increases in age-related
currency fluctuations, nondebt financing of deficits, and government spending.
accumulation of financial assets can all temporarily alter
the one-to-one link between debt and deficits, but in Approach 1: Convergence in the Long Term
general the debt path closely follows that of the deficit. This approach derives the constant fiscal balance19
In the same way, deficits are inherently tied to govern- that, if maintained, would lead to a gradual con-
ment spending, with cuts to the deficit often requiring vergence toward the debt target in the long term.
cuts to spending. As a result, any target set on the debt Equation (1) lays out the basic formula (derived in
path implicitly puts constraints on the deficit and, Appendix 2) that could be used for the calibration of
ultimately, on spending. Therefore, one needs to ensure both the overall balance and the primary balance tar-
consistency between the debt rule and the operational gets as a share of GDP (b *), for a given debt-to-GDP
rules placed on the deficit and spending. target (d *) and parameter λ:
In the methods presented below, the time horizon is
b* 5 d* (1)
generally longer than the time horizon used to calibrate
− γ
the debt ceiling in the previous section (which was six Note that λ is alternatively equal to ___ in the case
i − γ 1 + γ
years by default). Using identical time horizons in both of an overall balance target, and ___
1 + γ in the case of a
exercises would be unwarranted and potentially mislead- primary balance target, where γ stands for the nominal
ing. The calibration of the debt ceiling is based on the GDP growth over the long term and iis the nominal
prudence principle that debt dynamics should remain interest rate paid on public debt.20
under control even if negative shocks occur repeatedly Equation (1) suggests that, for a hypothetical
over the medium term. It would not be reasonable to country that grows by 5 percent in nominal terms and
extend this time horizon beyond five or six years. A pays 3 percent of nominal interest rate on its debt over
scenario of repeated negative shocks over the long term the long term, a 60 percent of GDP debt target would
(say, 20 years) would not only be unrealistic but would imply an overall deficit target of about 2.9 percent of
also certainly result in a safe debt level equal to zero. GDP (as shown in Figure 6) or a primary deficit target
On the contrary, the time horizon used to calibrate the of about 1.1 percent of GDP.
operational rules is a policy decision reflecting national This is the standard approach to calibrating the
preferences. The question is whether a government budget balance rule from a given debt target. It has,
wishes to attain the safe debt target asymptotically in the for example, largely inspired the calibration of the
long term or over a shorter horizon. European Union’s framework of fiscal rules. Depending
on the debt target and the initial fiscal balance, this
formula can entail either an instantaneous consoli-
From the Public Debt Ceiling to the Deficit Ceiling18 dation or expansion of the fiscal position. A positive
The derivation of the deficit ceiling from the debt feature of this approach is that it is simple and compels
ceiling can be done flexibly, depending on the timing an immediate adjustment of fiscal policy toward the
and sequencing of the desired adjustment path. In debt target. On the downside, convergence to the debt
what follows we will discuss four main approaches, target can be very slow. For example, it would take
each adding a layer of flexibility: (1) a constant balance about 15 years for our hypothetical country’s debt to
target that guides debt to its ceiling in the long term, complete half the distance from an initial debt ratio of
(2) a constant balance target that guides debt to its 70 percent to a target of 60 percent. In addition, this
ceiling by a given date, (3) a constant balance target approach relies on the simplifying assumption that λ is
that guides debt to its ceiling by a given date following
a transition period, and (4) a constant balance target 19In the rest of this section, the formulas are based on fiscal bal-
that guides debt to its ceiling by a given date, while ances. The results should be interpreted as a deficit ceiling or a deficit
target when the balance is negative.
20Growth and interest rates could be replaced by their real-term
18In countries with exhaustible natural resources, the framework values when deriving λin equation (1), under the simplifying
would need to be modified to take into account future resource rev- assumptions that nominal interest rates and nominal GDP are
enues (Baunsgaard and others 2012; IMF 2012). In this context, the deflated with a similar deflator and that there is no difference
PIH, Modified PIH, and Fiscal Sustainability Frameworks provide between actual and expected inflation. Note that the derivations
benchmarks to calibrate the threshold of fiscal balance rule and, indi- implicitly assume that either the share of foreign currency debt is
rectly, estimates of the sustainable level of expenditure. low, or the effective exchange rate is largely stable over the long term.
–2%
–3%
–5%
–6%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
2. Debt Path
74%
Approach 3:
72% Initial transition period
70%
Approach 1:
68% Convergence in the long run
66%
64%
62% Approach 4:
Buildup of fiscal buffers
60%
58% Approach 2:
Convergence by a given date
56%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Source: IMF Staff.
Assumptions: nominal growth, 5%; interest rate, 3%; initial debt, 70%; debt target, 60%;
initial balance, –5%.
constant over the long term; in reality, growth and the debt service. In contrast, a constant overall balance
interest–growth rate differential vary with the level of rule would place the debt on a convergent path: if the
public debt (IMF 2010, 2013c). overall balance is set at the constant level b *, the debt
The balance target derived in this approach is also ratio will asymptotically converge to d *from any initial
known as the “debt stabilizing fiscal balance,” which, in debt level (under the assumption that nominal growth
the case of the primary balance, is somewhat of a mis- γis positive).
nomer. Indeed, constant primary balance rules imply
explosive debt paths (that is, diverging to +/-∞), except Approach 2: Convergence by a Given Date
when (1) GDP grows faster than the interest rate paid This approach is similar to the first one, except that
on debt or (2) the starting debt level is already at the the balance rule (b *) is calibrated so that the debt ratio
target d0 = d *(Escolano 2010). Consider the case hits its target (d *) after N
years. Equation (2) below
in which growth is below the interest rate and the lays out the basic formula in this approach given the
initial debt ratio is above the target. In this case, the initial level of debt d 0, λ, N, and d *(see Appendix 2
increase in nominal debt (owing to the financing of for details).
interest payments) would be larger than the increase in
b* 5 __________
[d0 (1 1 )N 2 d*] (2)
nominal GDP. This would raise the debt-to-GDP ratio N
(1 1 ) 2 1
indefinitely, as the primary balance is kept constant at
the target ratio and does not adjust to offset the rising
bt 5
aT 1 b0 5 b * , when T t N
T
the Buildup of Fiscal Buffers
This approach offers another layer of flexibility in
the calibration of the balance rule by allowing for a
balance path that accommodates expected increases
in future spending. It is recommended for countries
21Note however that, in this approach, the derived balance
facing the prospect of aging costs but could also be
target will need to be adjusted to the debt-stabilizing balance after
the debt ratio reaches its target. Otherwise, the debt ratio would suitable with regard to other long-term costs, such
keep diverging. as those related to the environment. In an economy
22A linear adjustment schedule is not crucial to the outcome; what
with an aging population, targeting the constant fiscal
matters is the total adjustment αTat the end of year T that brings
the balance to the target. balance derived in the second approach (equation 2)
b*, when 0 , t , P (5)
sented in Figure 6, the debt ratio declines below target
bt 5
b* 2 At, when P t N
in the later years to accommodate future increases in
deficits that will bring debt back to target by the end
where ∆ At = A t − A 0 denotes the incremental aging of the forecast horizon. Note that this approach is usu-
costs in a given year, meaning the difference between ally suited to a much longer term perspective (decades)
total age-related costs A trelative to their size in the than the ones discussed above. We have used the same
initial period A 0 = A P−1.24 time horizon in Figure 6 (N = 15years) for ease of
Equation (6) shows the main formula deriving the comparison across approaches.
constant fiscal balance in this approach using similar A broadly similar approach is also used to calibrate
notations as above, given d 0, λ, N, dN * , and ∆ At (see the Medium-Term Objectives (MTO) of the Euro-
Appendix 2 for details). Note that equation (6) is an pean fiscal framework (European Commission 2016).
adaptation of the long-term adjustment need formula In particular, the MTO formula includes a term that
of the European Commission, also known as the “S2 covers a fraction of the present value of the projected
indicator” (European Commission 2015). increase in age-related expenditure.
From the Structural Deficit Ceiling to the Nominal (see Fedelino and others 2009 and Escolano 2010
Deficit Ceiling for a detailed discussion). For a government targeting
The previous formulas are used to ensure that the a structural deficit of 1 percent of potential GDP,
debt and balance rules are consistent in the long term; spending 40 percent of GDP on average, and facing an
that is, when nominal aggregates are equal to their output gap of –2 percent, equation (8) would imply a
steady-state structural values. It is thus reasonable to constant nominal deficit ceiling of 2 percent of GDP.
assume that these formulas calibrate the structural This formula is, for instance, used to ensure
balance rule.27 consistency between the nominal deficit ceiling and
It can be useful to derive from the structural balance the Medium-Term Objective (in structural terms)
rule a corresponding threshold for the nominal balance in the European fiscal framework (European Com-
rule. This is particularly important in countries that mission 2016).
have both nominal and structural balance rules, to
minimize the risk of conflict between them. For a From the Structural Balance Ceiling to the
given structural balance target, the nominal deficit Expenditure Ceiling
ceiling should be sufficiently high to allow automatic
stabilizers to fully operate during a typical economic The final step in deriving operational rules is the
downturn (this means that the nominal deficit should calibration of expenditure rules, which can be defined
be allowed to increase in response to the cyclical in terms of either expenditure growth or an expen-
decline in revenues). But the nominal deficit ceiling diture ratio. For simplicity, we assume that there
should not be so high that it permits discretionary fis- is no cyclical component to expenditure, meaning
cal expansions that are inconsistent with the structural that expenditure does not respond automatically to
balance ceiling. economic conditions (as would be the case if there
Equation (7) details how to compute a constant were an unemployment benefit scheme, for instance)
nominal balance target, nb *, for a given structural bal- and that automatic stabilizers operate on the revenue
*(See Appendix 2 for its derivation).
ance target, sb side only. This is consistent with empirical evidence
showing that revenues are far more sensitive than
nb* 5 sb* 2 OGmax[r(1 2 ) 2 e(1 2 )] (7) expenditure to the business cycle (Price, Dang, and
where OGmaxstands for the maximum output gap Guillemette 2014). In this case, there is no difference
during a typical downturn (it is a negative number); r between observed/nominal expenditure and structural
and ηare, respectively, the revenue ratio and its elas- expenditure. We also assume that the country has a
ticity relative to output; and eand κ are, respectively, given structural tax ratio r s(computed as the ratio of
the spending ratio and its elasticity relative to output. structural revenues to potential GDP), which remains
Estimates of revenue and spending elasticities gen- constant unless there is a change in tax policy.
erally yield values close to 1 and 0, respectively; hence Under these assumptions, for a given structural
the following is a common proxy for the link between balance ratio sb, the implied expenditure ratio e s (in
structural and nominal balances: percentage of potential GDP) is equal to:
The structural balance target sb *is expressed as a which implies that:
ratio to potential output while the nominal balance e s 5 r s 2 sb (9)
target nb *is expressed as a ratio to current output
If there is no tax policy change (∆ rs = 0) and if the
27In principle, the deficit ceilings derived from the various country already complies with the structural balance
approaches discussed here could be applied to either the nominal rule (sb = sb * and ∆ sb= 0), the two equations show
balance or the structural balance. When applied to the structural bal-
that the structural balance rule can be interpreted as
ance, the budget balance ceiling is binding on average over the cycle
(although the nominal balance can deviate and fluctuate with the (1) a constant ratio of spending-to-potential GDP
cycle). When applied to the nominal balance, however, the ceiling is ∆ es = 0or (2) a rule in which spending growth is
constantly binding and must be enforced every year. This means that equal to potential GDP growth. In other words, if the
the fiscal deficit consistent with the debt target is more demanding
when the formulas are applied to the nominal balance than to the country is already at a structural position consistent
structural balance. with the structural balance rule, nominal spending
should grow at the same pace as nominal potential the expenditure rule is key to preserving incen-
GDP. tives for taking new revenue-enhancing measures
For instance, for a country that is collecting 40 per- as well as to ensuring that the adoption of new
cent of potential GDP in revenues and whose poten- tax expenditures (or other revenue-decreasing
tial GDP is growing in nominal terms by 4 percent measures) does not derail the debt path.
a year on average, maintaining a structural deficit of •• The framework can also be adapted to fit transi-
1 percent of potential GDP would be consistent with tion periods toward the structural balance target
28Another option is to combine the expenditure rule with a deviations from the debt path (see Debrun, Epstein, and Symansky
debt-break mechanism that forces a correction of the rule for past 2008 for a detailed discussion of the case of Israel).
Appendix 1. Deriving the Debt Rule Threshold roeconomic variables. N sequences of six-year projec-
tions can be obtained by drawing repeatedly from this
Method One distribution.
A multivariate normal distribution of a
This method requires the characterization of the
k-dimensional vector of macroeconomic variables can
joint distribution of the macroeconomic variables
be written as
needed to project the public debt ratio. For advanced
and emerging market economies, these variables are x Nk (, )
growth, the average interest rate on debt, and the
with the k‑dimensional mean vector
exchange rate. For low-income countries, the terms
of trade and external financing disbursements are 5 (E[X1], E[X2], . . . , E[Xk])
also required. Characterizing the joint distribution is
and the k × kcovariance matrix
done either by (1) estimating a vector autoregression
(VAR) or (2) directly calibrating a joint multivariate 5 (cov(Xi, Xj)), for all i 5 1,2, . . . . , k;
distribution based on historical co‑movement between j 5 1,2, . . . , . k
The parameters μ
variables. A VAR can be used when quarterly data are
, Σcan be calibrated based on the
available, so that there are sufficient observations for
historical mean, variance, and covariance of macro-
econometric estimation. The direct calibration of a
economic variables. Alternatively, the econometric
joint multivariate distribution may be most appropri-
programs accompanying this note allow for the use
ate where only annual data are available.
of a multivariate Student’s t distribution, which has
Indirect Estimation with a VAR wider tails. (See manuals describing the files for more
information.)
If quarterly data are available, an unrestricted VAR
can be estimated for each country, describing the joint Calibrating the Debt Ceiling
dynamics of the macroeconomic variables needed to The debt ceiling is calibrated as follows:
project public debt. The econometric model is
Xt 5 A0 1 Aj Xt 2 j 1 �t
p 1. A set of macroeconomic variables is forecast over
a six-year projection horizon N times by either (a)
j51 using the estimated VAR, including shocks each
where Xt is the k-dimensional vector of macroeconomic period; or alternatively by (b) drawing directly from
variables and ε t is a k‑dimensional vector of normally the calibrated multivariate distribution of macroeco-
distributed shocks: ε t ∼ N(0, Ω). nomic variables each year.
The estimated variance-covariance matrix of the 2. The N sets of macroeconomic variable forecasts
VAR, Ωˆ is then used to generate N sequences of are used to generate N trajectories of the primary
macroeconomic shocks ε tover the six-year projection balance, using a fiscal reaction function (FRF)
horizon (N is a large number of simulations; for exam- and the previous year level of debt (see Box 3 in
ple, more than 1,000). For each of the N simulations the text for FRF options). In the econometric
of shocks, the estimated VAR model is used to forecast programs accompanying this note, annual changes
macroeconomic variables X tover the six-year projec- in the primary balance implied by the FRF are
tion horizon, adding the generated shocks to the VAR constrained (that is, they cannot exceed certain
model each year as the error term. Using a VAR to limits) on the basis of historical experience to
make projections is ideal, as the lagged effect of mac- ensure that projected primary balances are realistic.
roeconomic shocks can be taken into account through Fiscal shocks can also be added directly in the
the autoregressive structure of the model. FRF. The distribution of fiscal shocks is calibrated
based on estimated deviations between actual fiscal
Direct Calibration of a Multivariate Distribution responses observed (that is, actual levels of the
As an alternative to VAR estimation, a simpler primary balance) and the fiscal response predicted
approach can be used if quarterly macroeconomic data by the FRF within the sample.
are not available. A multivariate normal (or Student’s 3. The N corresponding trajectories of debt (starting
t) distribution of key macroeconomic variables can be at the current debt level) are obtained by the system
calibrated based on historical comovements of mac- of simultaneous equations formed by the debt accu-
( (
rt 2 gt
dt 5 1 1 _____
1 1 gt ))
dt 2 1 2 pbt 1 SFAt 1. A set of macroeconomic variables is forecast
over a six-year projection horizon N times
where dt is debt (as a ratio of GDP), r t is the aver- using an estimated VAR, including shocks each
age effective real interest rate on debt, g t is the real period (the VAR estimation is similar to that in
GDP growth rate, pb tis the primary balance (as a Method One).
ratio of GDP) and SFA tis the stock‑flow adjust- 2. The N sets of forecasts are used to generate N
ment (as a ratio of GDP). The debt accumulation trajectories of the primary balance, using either an
equation includes a constant stock flow adjustment estimated or normative fiscal reaction function (see
each period that could potentially account for reali- Box 3 in the text).
zation of contingent liabilities. 3. The N corresponding trajectories of debt (starting
4. If the 95th debt percentile (or other chosen percen- at the current debt level) are obtained by the system
tile, given risk tolerance) of the debt ratio distribu- of simultaneous equations formed by the debt accu-
tion is significantly below the maximum debt limit mulation equation (government budget constraint)
(MDL) in all years of the projection horizon [or is and the fiscal reaction function (which depends on
significantly above the MDL in at least one year], the lagged value of debt).
the starting level of debt is increased [decreased] 4. If the 95th percentile of primary balances (or
by a small amount (0.3 percent), and steps 1–3 are other chosen percentile, given risk tolerance) is
repeated based on the new starting level. significantly below the maximum feasible primary
surplus (MFPS) (that, is, falls below the interval
Steps 1–4 are repeated until the 95th percentile of pb 95 ∈ [MFPS − 0.4; MFPS + 0.4]) in all years of
the debt level falls into a small interval around the the projection horizon [or is significantly above
MDL in at least one year of the medium-term pro- the MFPS in any one year], the starting level of
jection horizon: Debt
95 ∈
[MDL − 0.4; MDL + 0.4], debt is increased [decreased] by 0.3 percent, and
without significantly breaching the MDL in any year. steps 1–3 are repeated based on the new starting
The starting level of debt satisfying this criterion is level. For advanced economies a typical MFPS
called the debt ceiling; that is, the level of debt from would be about 4 percent of GDP, while for emerg-
which its projection does not exceed the MDL with ing market economies it would be 2 percent of
95 percent likelihood over the medium-term projec- GDP (see Escolano and others 2014).
tion horizon. The safety margin is computed as the
MDL minus the debt ceiling. Steps 1–4 are repeated until the 95th percen-
The fan charts can also be used to determine the tile of the primary balance falls into the small
probability of breaching the maximum debt limit, interval around the MFPS in at least one year
conditional on any starting level. For example, using of the medium-term projection horizon:pb 95 ∈
the current debt level as the starting level, the fan chart [MFPS − 0.4; MFPS + 0.4], without significantly
can be used to determine the probability that debt will breaching the MFPS in any year. The starting
exceed the debt limit in all years over the projection level of debt satisfying this criterion is called the
horizon. This can be useful to gauge the extent of risk debt ceiling.
associated with a country’s current debt position.
We start with the accounting identity relating the Using equation (A.2.2), the instantaneous adjust-
gross debt to overall balance (OB) and primary balance ment to the balance ratio to reach a constant debt ratio
(PB), also known as the debt dynamics equation:1 d *over the long term can be decomposed as
1
Dt 5 Dt 2 1 2 OBt ⇒ dt 5 ______ dt 2 1 2 obt
b* 2 b0 5 [d0 2 b0] 2 (d0 2 d*)
(1 1 t )
(1 1 it ) where the first term on the right-hand side corresponds
Dt 5 (1 1 it ) Dt 2 1 2 PBt ⇒ dt 5 ______ d
(1 1 t ) t 2 1 to the gap to the debt-stabilizing balance ratio (at the
2 pbt
initial level of debt), and the second term corresponds
where Dt is the nominal level of debt at the end of to the additional adjustment due to the debt target d *
period t, it is the effective nominal interest rate paid on being (potentially) different from the initial debt ratio
the inherited debt D t−1, and γ tis the nominal growth d 0.2 While the pace of adjustment is instantaneous in
rate of GDP, with small letters denoting a variable this particular example, the convergence to the desired
as a share of GDP. We can rewrite this debt dynam- debt ratio would be asymptotic.
ics equation as
Approach 2: Convergence by a Given Date
dt 5 ( 1 1 jt ) dt 2 1 2 b tj for j {o,p}
Under the assumption that λ t is time-varying, equa-
1
______ , when b tj 5 obt for j 5 o tion (A.2.1) has the following solution
(1 1 t )
N
______
j N N
1 1 t 5
(1 1 it )
(1 1 t ) , when b tj 5 pbt for j 5 p
dN 5 d0 (1 1 t ) 2
t51
t51 i5t11
(1 1 i ) bt
Assuming that λ tis time-invariant, the solution to
with bt j denoting the balance ratio in period t, which
equation (A.2.1) becomes
becomes the overall balance if j = oand primary
balance if j = ptogether with the corresponding dN 5 d0 (1 1 )N 2 tN5 1 (1 1 )N 2 t bt (A.2.3)
− γ t
growth-adjusted interest rate λ ot = _____
if j = o and Given the initial debt level d 0, the constant balance
1 + γ t)
(
it − γ t
_____
λ t =
p
if j = p. In the rest of the appendix, we ratio b *that achieves a target debt ratio d N* in N
peri-
( 1 + γ t)
will drop the index jand derive the relevant equations ods N < ∞can be derived from (A.2.3) as
for the balance ratio; one could retrieve the relevant b* 5 _____________
1
(d0 (1 1 )N 2 d * )
N N2t
t 5 1 (1 1 ) N
equations for the overall and primary balances using
corresponding λ jt for j = {o, b}: which leads to
dt 5 (1 1 t ) dt 2 1 2 bt (A.2.1) b* 5
__________
*N ) (A.2.4)
(d0(1 1 )N 2 d
(1 1 )N 2 1
Rearranging equation (A.2.4) to decompose the
Approach 1: Convergence in the Long Term components of adjustment as before, we get
d0 2 d *
Using equation (A.2.1) and assuming λ t = λ is b* 2 b0 5 [d0 2 b0] 1 ____________
N
(A.2.5)
N ((1 1 ) 2 1)/
constant in the long term, the constant balance ratio b *
would give b * − b0 = [λ d0 − b 0] + [ d0 − d N* ] , where the
compatible with achieving a constant debt ratio d * in Setting N = 1in equation (A.2.5), for example,
the long term would be
instantaneous adjustment to the balance ratio required
b* 5 d* (A.2.2) to bring the debt ratio to d *in 1period would be the
Specifically, the constant overall and primary bal- sum of (1) the gap to the debt-stabilizing balance ratio
ance ratios are (at the initial debt level) and (2) an additional term
2Note that when λ < 0 and d * < d0 , the second term is positive,
implying additional adjustment relative to the adjustment needed to
stabilize the debt at its initial level. However, if λ > 0, the debt path
1If gross debt becomes negative, it will be interpreted as assets. becomes unstable, diverging to + ∞if d * < d0 and − ∞ if d * > d0 .
reflecting the distance between the initial and desired achieve dN * , which could be very costly. As T
increases,
debt ratios. the yearly adjustment α falls:4
( )
*
d0 2 d
d0 2 b0 1 ____________
N
Approach 3: Convergence by a Given Date Following ((1 1 )N 2 1)/
____________________________________
(
a 5
a Transition Period
)
T(1 1 )N 2 T ((1 1 )T 2 1)/ 2 tT5 1 (1 1 )N 2 tt
T 2 _____________________________________
This α(T, b0 , N, λ, d0 , dN * )pins down the path for the
((1 1 )N 2 1)/
Let the initial balance ratio b 0 in period t = 0be
gradually adjusted until period T < N, and achieve
b T* in period T, which is the constant balance ratio balance ratio as depicted in (A.2.6), and using (A.2.7)
that brings the debt ratio to d N* by period N . Assum- we would have
I
ing a linear adjustment schedule for the balance b T* 5 aT 1 b0 5 __________
N
[d0(1 1 )N 2 d N*
(1 1 ) 2 1
ratio would give
at 1 b0, when 0 t T 1 A(T, b0, N, , d0, d N* )]
where A(T, b0 , N, λ, d0 , dN * ) = α ∑ Tt=1 (1 + λ) N−t(T − t).
bt 5 (A.2.6)
aT 1 b0 5 b*T , when T t N
with αdenoting the annual adjustment of the fiscal Approach 4: Convergence by a Given Date Following
balance from period 0to T. In this case, equation a Buildup of Fiscal Buffers
(A.2.3) becomes
T Let ∆ At = A t − A 0be the change in age-related
dN 5 d0 (1 1 )N 2 (1 1 )N 2 t (at 1 b0) spending relative to period t 0, and b *be the constant
t51
N balance ratio excluding the incremental change in
2
(1 1 )N 2 t (aT 1 b0)
age-related spending ∆ A tthat, if maintained, would
t5T11
ensure convergence to the debt target d N* by the end
Rearranging3 to decompose the components of the of period N .5 In other words, the balance ratio fol-
b0 = αTleads to
total adjustment b T* − lows this path:
d0 ((1 1 )N 2 1)
_____________ bt 5 b* 2 ∆ At (A.2.8)
aT 5
N
N2t 2 b
t 5 1 (1 1 ) 0
In this case, equation (A.2.3) becomes
aT (1 1 )N 2 t(T 2 t) d 2 d *
1 ___________________
t5 1 1 _____________
0 N N
I II
d0 2 d * where S = ∑ N (1 + λ) N−t ∆ At is the value in year N
of
1 ____________ (A.2.7)
N t=1
N
((1 1 ) 2 1)/ the cumulative increase in long-term age-related costs
through N. Similar to (A.2.7) we could decompose
III
S
1 ____________
N
the elasticities of revenue and expenditures with respect
((1 1 ) 2 1)/
to the output, which measure the relationship between
IV
the cyclical components of revenues and expenditure
where (IV) is the additional required upfront adjust-
relative to the cyclical component of output. In this
ment due to costs of an aging population.
case, the structural balance as a ratio of potential GDP
Note that if we impose the simplifying assumption
(sb) is defined as
that age-related costs increase linearly by an annual
amount δstarting from year P t = δ(
≤ N, i.e., A
SB
sb 5 __s Y
R
Y
s
E s
Y
RR
Y R
s
E E
Y E
s R Y
Y ( Y )
5 __s 2 __s 5 _s __ 2 __s __ 5 __s __
s
t + 1 − P) + A 0 for t ≥ P, the analytical solution for S
can be derived as
E __
2 __ ( )
Y s
Y s Y
S 5 (1 1 )(N 1 1 2 (P 2 1))/2 R
5 _________
1
Y/(1 1 OG ) ( 1 1 OG )
______
E
2 _________
1
Y/(1 1 OG ) ( 1 1 OG )
______
k
1 2 __________________
1 1 (N 1 1 2 (P 2 1))
(1 1 )
N 1 1 2 (P 2 1)
R
5 __ E
(1 1 OG )1 2 2 _ (1 1 OG )1 2 (A.2.10)
Y Y
Under the special case in our simulations, in which
For a small enough O G, one can approximate
A t = A 0 until year P − 1and starts increasing from year
(1 + OG) ≈ 1 + (1 − x)OG for x = {η, κ}.
1−x
P, the balance ratio would remain constant at b *as
In this case, equation (A.2.10) becomes
given by (A.2.9) until year P − 1then would gradually
decline by ∆ A tstarting from year P as given by (A.2.8). sb 5 r (1 1 (1 2 )OG ) 2 e(1 1 (1 2 )OG )
5 (r 2 e) 1 OG[r (1 2 ) 2 e (1 2 )]
From the Structural Deficit Ceiling to the Nominal
Deficit Ceiling 5 nb 1 OG[r (1 2 ) 2 e (1 2 )] (A.2.11)
Let Y and Y sdenote the actual and potential nom-
where rand e respectively denote revenue and expen-
inal GDP, R and R sthe levels of actual and structural
ditures as a share of GDP, with the nominal fiscal
revenues, E and E sthe levels of actual and structural
balance (nb) as a share of GDP given by n b = r − e.
expenditures, and N B = R − Eand S B = R s − E s the
Using (A.2.11) and taking a medium-term view
Y − Y be s
nominal and structural balance. Let OG = ____
s (averages of revenue and expenditure ratios), one
Y