documento de trab - economía uceconomia.uc.cl › wp-content › uploads › 2018 › 08 ›...
TRANSCRIPT
D O C U M E N T O
D E T R A B A J O
Instituto de EconomíaD
OC
UM
EN
TO d
e TR
AB
AJO
I N S T I T U T O D E E C O N O M Í A
www.economia.puc.cl • ISSN (edición impresa) 0716-7334 • ISSN (edición electrónica) 0717-7593
Fiscal Rules and Fiscal Performance: World Evidence
Klaus Schmidt-Hebbl y Raimundo Soto.
5172018
1
Fiscal Rules and Fiscal Performance: World Evidence *
Klaus Schmidt-Hebbel **
Raimundo Soto ***
Final Version, September 2017
Abstract
This paper presents world evidence on the contribution of fiscal rules to fiscal performance.
The paper identifies the links between fiscal institutions, rules, and performance measures,
and reviews the relevant literature on fiscal performance and the potential contribution of
different fiscal rules to performance. The expected contribution of rules to stabilization
objectives is ambiguous, depending on policymakers´ objective functions and constraints, as
well as on particular design and implementation features of rules. Rules have expected effects
on other policy objectives (solvency and government size) that are less ambiguous. This
paper’s empirical models identify the potential contribution of three types of rules –
expenditure, budget balance, and debt rules, using de facto and de jure measures – to four
indicators of fiscal performance, controlling for other determinants. The empirical analysis is
conducted using an econometric approach designed to deal with potential endogeneity,
dynamic responses, and unobserved heterogeneity. The models are tested on a world sample
for 115 countries, spanning 1985-2015. De jure expenditure rules (but not other rules) are
found to reduce expenditure procyclicality. Yet procyclicality of budget balances is not
affected by any rule. All three types of rules raise government budget balances but they have
no significant effects on government debt. The latter results are broadly robust to using
alternative de facto rule measures. The results are also generally robust to nested testing of
differential effects in country sub-groups, i.e., no systematic effects of rules on fiscal
performance were found for small states and for Latin American countries.
* Prepared for the World Bank LCR Regional Study on Fiscal Rules and Economic Size. We
thank Fernando Blanco, Fritzi Koehler-Geib, Emilia Skrok, and Carlos Végh for helpful
discussions and comments provided to an earlier draft. We also thank Isaac Martinez for very
able research assistance. The usual disclaimer applies.
** Instituto de Economía, P. Universidad Católica de Chile, email: [email protected]
*** Instituto de Economía, P. Universidad Católica de Chile, email: [email protected]
1
1 Introduction
Fiscal rules and restrictions that govern the conduct of fiscal policy have been in place
in a small number of industrial countries for a long time. However, only since the 1990s
adoption of fiscal rules started spreading world-wide, as part of significant reforms of fiscal
frameworks in many industrial and emerging/developing countries (Figure 1).
Figure 1 Number of Countries with Fiscal Rules in Place, 1985-2015
Source: IMF Fiscal Rules Dataset, 2016.
Reforms of fiscal institutions and fiscal rules have been motivated by several (often
complementary) objectives: (i) making fiscal policy design and execution more resilient to
principal-agent and other political-economy problems that cause deficit bias, (ii)
strengthening fiscal solvency and sustainability (i.e., attaining sustainable levels of
government deficits and public debt), (iii) contributing to macroeconomic or cyclical
stabilization, by reducing fiscal policy pro-cyclicality or raising fiscal policy counter-
cyclicality, and (iv) improving intergenerational fairness (IMF, 2013).
The latter objectives are shared by most fiscal policy makers worldwide. So why do
some countries adopt fiscal rules while others do not? Recent research has shed light on
political, institutional, and economic conditions that explain the likelihood of adopting fiscal
rules and maintaining them over time (Elbadawi et al. 2015, Schmidt-Hebbel and Soto 2017).
2
This paper focuses on the original motivation of policy makers when putting in place a
fiscal rule: the attainment of several fiscal policy objectives. There is little empirical research
available on the possible contribution of fiscal rules to the attainment of fiscal objectives. This
study fills this void.
Of the four policy objectives identified above, this study will focus on the contribution
of fiscal rules on two objectives: fiscal solvency and sustainability, and cyclical stabilization. It
will do so by identifying the statistical contribution of fiscal rules to four different empirical
measures of fiscal performance that are proxies for the policy objectives of solvency and
sustainability, and cyclical stabilization. Countries have adopted four types of fiscal rules (as
depicted in Figure 1): rules on government budget, debt, expenditure, and/or revenue. There
are also two different measures of fiscal rules available: legal or de jure rules and measures of
rule compliance or de facto rules (using the recent dataset on fiscal rules compliance by
Belinga et al. 2017).
Specifically, this paper will address the following questions:
(1) Do fiscal rules promote good fiscal performance? (2) Does the type of fiscal rule matter affect for its impact on fiscal performance? (3) Which other fiscal institutions, political institutions, macroeconomic regimes,
development conditions, and macroeconomic conditions affect fiscal performance and sustainability?
(4) Which factors matter for the effectiveness of de jure fiscal rules? I.e., do fiscal rules interact with other institutional and economic variables in their impact on fiscal performance?
(5) Does the effect of fiscal rules on fiscal performance depend on country size? (6) Is there a difference between the impact of de jure and de facto fiscal rules?
The paper’s next section discusses the multi-causal links between fiscal institutions
(including rules) and fiscal performance, defines measures of fiscal performance and types of
fiscal rules, reviews selectively the relevant literature on optimal fiscal policy, derives a policy
function from government optimization in a small open economy, and presents a taxonomy of
fiscal rules and their relation to fiscal policy objectives. Section 3 reviews selectively the
empirical evidence on the determinants of fiscal performance and the contribution of fiscal
rules to fiscal performance, as well as the reverse causality from performance to rules. Section
4 addresses empirically the question if small economies are structurally different from larger
economies, which is an important issue for the subsequent empirical research. The empirical
methodology is discussed in section 5, addressing the main econometric issues, including the
key issue of potential endogeneity of fiscal rules to fiscal performance. The subsequent
empirical research is conducted using a three-stage instrumental-variable approach designed
to deal with the challenges of assessing the impact of fiscal rules: potential endogeneity,
dynamic response, and unobserved heterogeneity. Section 6 deals with the creation of
suitable instruments for a variety fiscal rules, as well as for other fiscal institutions, such as
sovereign wealth funds. Section 7 presents the core research results on the contribution of
different types of de jure fiscal rules to the four measures of fiscal performance, controlling for
a relevant number of other performance determinants. Section 8 reports comparable results
for the contribution of de facto rules to fiscal performance. Section 9 concludes.
3
2 Fiscal Rules and Fiscal Performance: Measures, Interactions, and
Policy Optimality
A fiscal policy rule is one among several components of a fiscal framework adopted to
strengthen fiscal performance. How is fiscal performance linked to fiscal institutions
(including rules), considering country conditions? How do we define fiscal performance and
which are the main types of fiscal rules? How are optimal policy functions (not fiscal rules)
derived in the analytical literature? Which are the expected effects of different sub-categories
of fiscal rules on different fiscal performance measures? We address these questions in this
section.
2.1 Interactions between Fiscal Performance, Fiscal Institutions, Economic Conditions, and Political-Economy Features
Fiscal institutions and fiscal performance measures are linked in several causal
directions and affected by other variables and conditions. These complex multi-factor and
multi-causal relations are depicted in Diagram 1. Fiscal performance (measured by debt and
deficit levels, public debt prices or yields, and policy cyclicality) is influenced by fiscal
institutions (including rules, independent fiscal councils or agencies, and sovereign wealth
funds), cyclical and structural conditions (business cycles, economic shocks, per capita GDP,
natural resources, and country size, among others), political-economy distortions (caused by
principal-agent problems, political cycles, high discount rates, interest groups, etc.), and
political institutions (democracy, institutional quality, checks and balances, electoral rules,
among others).1
Reverse causality from fiscal performance to fiscal institutions is observed in the
design, adoption, and functioning of fiscal institutions, which are all shaped by fiscal
performance measures and other economic and political determinants.
2.2 Measures of Fiscal Performance
Fiscal policy seeks attainment of several different objectives, ranging from
macroeconomic goals (fiscal sustainability and solvency, output stability, and size of
government), to microeconomic and sector goals (more efficiency in resource allocation
across sectors and regions, removing distortions, and supplying public goods), and to social
goals (reducing poverty and raising equity). The structure of public expenditure and taxation
is geared to attaining micro, sector, and social goals. In contrast, setting aggregate levels of
government spending, revenue (including taxes), and/or debt levels are the conventional
instruments of choice to attain macroeconomic goals. We focus on the latter in this paper.
There are different approaches to assess directly a country’s fiscal sustainability and
solvency. Some measures are test results based on debt stationarity conditions (Milesi-
Ferretti and Razin, 1996), typically applied to one specific country, with detailed information
1 Here we include fiscal rules as one of several fiscal institutions, defined in a broad sense. In a narrow sense, fiscal institutions are government entities such as government agencies and funds. A fiscal framework comprises institutions (in a narrow sense), fiscal rules, and the legal and regulatory acts that define the latter and constrain fiscal policy decisions.
4
about its public finance structure and on detailed projections of key future macroeconomic
and financial variables that affect its fiscal stance. Considering the wide range of reasonable
assumptions and projections that underlie a fiscal sustainability study, it is not surprising that
different parties come to different conclusions regarding sustainable debt and deficit levels.
(E.g., the ongoing discussions of Greece’s fiscal sustainability among creditor institutions).
Diagram 1 Fiscal Performance, Institutions, and Other Conditions
Cyclical and Structural Conditions
- Shocks/business cycles
- Per capita income
- Natural resources
- Market failures and rigidities
- Country size
Fiscal Institutions
- Fiscal rules
- Fiscal responsibility acts (constitutional or legal)
- Independent fiscal councils / agencies
- Sovereign Wealth Funds
Fiscal Performance
- Deficit levels
- Debt levels
- Debt prices / premiums
- Policy cyclicality
Political Institutions
- Democracy
- Institutional quality
- Checks and balances
- Electoral rules
Political-Economy Distortions
- Principal-agent problems
- Political cycles
- High discount rates
- Interest groups
- Common-pool problems
- …
5
An alternative gauge of fiscal sustainability is derived from the public debt limit,
beyond which (without extraordinary measures) debt would be unsustainable. Ghosh et al.
(2013) propose estimation of the debt limit from the reaction function of a country’s primary
(or non-interest) government budget balance to past government debt levels. Rolling time-
series estimation of the reaction function could reveal time-varying estimates of the debt
limit. Fiscal space is the difference between the debt limit and current debt. As both the debt
limit and current debt change over time, fiscal space varies accordingly.
Market measures of government debt prices or yields (sovereign debt yields or CDS
spreads) are an alternative, market-based gauge of fiscal sustainaibility.
No systematic information for large panel datasets is available for estimates of
sustainable public-sector debt and deficit levels (based on estimates a la Milesi-Ferretti and
Razin) and for estimates of debt limits and fiscal space (a la Ghosh et al. 2013). The same is
true for sovereign debt yields and CDS spreads.
In the absence of the latter, we will use in this paper the conventional approach of
using GDP ratios of government budget balances and government balance levels as proxy
indicators of fiscal sustainability and solvency.
Regarding fiscal policy cyclicality, we focus on two measures: the correlation between
government expenditure and GDP, and the correlation between the government balance and
GDP. For the first (second) measure, a positive (negative) correlation reflects pro-cyclical
fiscal policy.2
2.3 Types of Fiscal Rules
As discussed above, fiscal rules (among other fiscal institutions) are put in place in
order to strengthen fiscal policy objectives, which are more difficult to attain in their absence
due to several market failures and frictions, and to political-economy constraints. In this
sense, fiscal rules are second-best policies that potentially make more likely attainment of
fiscal objectives.
Before turning to the types of fiscal rules considered in this paper, it is key to consider
three issues about fiscal rules that are not much addressed in the analytical and the policy
literature.
Difference between fiscal (target) rules and behavioral fiscal rules
A fiscal (target) rule sets a quantitative target for a fiscal aggregate (for example, the
budget balance).3 In contrast, a behavioral fiscal rule is a policy function that links a fiscal
target (for example, the budget balance target) to a fiscal policy instrument (for example,
government expenditure), controlling for other variables that affect the economy, the budget
balance, and the fiscal policymaker’s objective function. This is analogous to the conduct of
2 Both are ex-post measures of overall policy cyclicality, combining the cyclical stance of discretionary and non-discretionary government expenditure (and implicitly those of government revenue, in the case of fiscal deficits). 3 We follow in this paper the conventional usage of “fiscal rule” instead of the more precise term “fiscal target rule”. As discussed below, terminology of policy targets and behavioral policy functions is used correctly in the realm of monetary policy.
6
monetary policy, where a policy function for the policy instrument (for example, the interest
rate in a Taylor rule) is co-determined by a target rule (for example, the inflation target),
controlling for other variables that affect the economy.
The analytical and policy literature (reviewed below) derives fiscal policy rules. Of
course, policy rules are unobservable variables – they may be hypothesized about or inferred
from econometric estimation, but they are not measured or reported as observable time
series. However, fiscal rules are publicly available, and hence will be used in this paper.
On fiscal rules
In order to clarify interpretation of the subsequent analysis and empirical results, next
we highlight several key issues related to design and implementation of fiscal rules.
Different fiscal objectives can be attained by different target rules. For example, the goal
of fiscal sustainability can be achieved by a rule on either the budget balance or the level
of debt.
Sometimes two (or more) different rules are defined for the same objective. The two (or
more) targets may or may not be consistent.
Inconsistency between two (or more) rules is less likely when the rule is set
asymmetrically, i.e., when it is defined as either a ceiling or a floor for two (or more)
targets. For example, the EU Maastricht target rules for government debt and deficits are
defined as ceilings.
When fiscal indicators are not close to their target values, governments face fiscal policy
space. In this case, targets are not restrictive and do not constrain fiscal policy behavior.4
The relation between fiscal objectives and fiscal rules gets even more complicate when
considering that governments aim at different policy objectives and use different rules
simultaneously. For example, attaining fiscal sustainability, i.e., lowering debt by running
budget surpluses when current debt exceeds the debt target, constrains heavily the use of
government expenditure as a counter-cyclical policy tool, at least during the transition
period toward lower debt. We illustrate this issue deriving a simple optimal fiscal policy
rule in section 2.5 below.
For any rule, the attainment of the fiscal objective depends on how the specific rule is
defined. For example, a budget balance rule has very different implications for the
cyclicality of fiscal policy, depending if it is defined on a year-by-year basis or over the
business cycle. We discuss this issue in more detail in section 2.6 below.
A final relevant issue refers to the government level for which a fiscal rule is defined.
Supranational rules are internationally agreed rules among members of an international
agreement or an economic union. National rules are adopted by individual countries as a
results of their sovereign decision or in obedience of a supranational rule. Separate fiscal
rules are adopted at sub-national levels (at state, provincial or municipality levels). We
limit our analysis below to national fiscal rules.
De jure and de facto rules
4 An alternative use of the term fiscal space is defined by the difference between a financially relevant limit and the actual level of a fiscal indicator, such as the debt limits defined by Ghosh et al. (2013).
7
Having a fiscal rule in place does not ensure compliance with the rule. One reason of
non-compliance with a fiscal rule is the existence of fiscal space. As discussed above, if a
government’s debt level falls significantly below its debt ceiling, it may decide to exceed
significantly its deficit target. On the other hand, governments often violate their rules – either
by deviating significantly from them or by exceeding significantly the corresponding ceilings –
as a result of economic or political restrictions. The widely documented, frequent violations of
deficit and debt rules by EU member countries are the best example of non-compliance with
de jure rules.
Types of fiscal rules
Now we turn to the fiscal rules used in this paper. We consider four types of fiscal
rules: on budget balance, government expenditure, government revenue, and government
debt. Our main focus in this paper is on measures of de jure rules. However, as an extension,
we will also consider measures of de facto rules.
Which are the links between particular fiscal rules and the main fiscal policy
objectives? Here we follow Schaechter et al. (2012). Debt rules and current budget balance
rules, and in particular when they are defined as debt and deficit ceilings, aim at achieving
fiscal (or debt) sustainability. Some forms of budget balance rules (structural balance rules),
expenditure rules, and revenue rules (those that limit the use of cyclical or natural resource
windfall revenue) aim at macroeconomic stabilization. Finally, expenditure rules (ceilings)
and revenue rules (floors) aim at limiting the size of government.
2.4 Optimal Fiscal Policy
The analytical literature on fiscal rules derives behavioral or policy functions from
optimization in an economy constrained by particular economic or political frictions. Some of
the behavioral fiscal rules are derived from the optimization of the government’s objective
function while others arise from the maximization of consumer welfare.
Leith and Wren-Lewis (2005) analyze optimal counter-cyclical fiscal policy, when
using tax and spending instruments, aimed at maximizing consumer welfare in an economy
with real and nominal frictions. The authors argue that political-economy distortions often
lead governments to mismanage short-term stabilization tools as an excuse to neglect long-
term sustainability. They argue that a well-designed explicit fiscal rule is an effective fiscal
institution to maintain fiscal sustainability.
Halac and Yared (2014) derive an optimal expenditure function for a government with
time-inconsistent preferences with a bias toward current spending. The government chooses
an expenditure rule as a trade-off between desire to commit to not overspend against its
desire to have flexibility to react to privately observed shocks to the value of spending. When
shocks are persistent and large, the rule does not preempt the government from deviating
from the ex-ante optimal rule, leading to erosion of fiscal discipline over time. This relevant
result provides an analytical foundation for the observed fact that fiscal rule enforcement may
be weak, leading to discrepancies between de jure and de facto rules when political-economy
frictions are present.
8
A different strand of the literature on optimal policy analyzes the interaction of
optimal fiscal and monetary rules in general equilibrium. Benigno and Woodford (2003)
derive optimal behavioral fiscal and monetary rules in an economy that is neoclassical in the
sense of fiscal policy (i.e., taxes are distortionary) and that is Neo-Keynesian in the sense that
prices and wages adjust slowly. Eser (2009) addresses similar interactions between optimal
fiscal and monetary policy for different benchmark economies.
The above cited literature is for closed economies. Végh (2013) extends the analysis of
optimal fiscal policy to open economies, focusing on the objective of macroeconomic
stabilization of a benevolent government that satisfies intertemporal consistency and
maximizes consumer welfare. In the benchmark case of full access to international capital
markets, no uncertainty, and no political-economy restrictions, full intertemporal
consumption smoothing is achieved by a government setting a constant consumption tax rate
(its policy instrument) over time. As a result, government spending (considered exogenous in
this model) is acyclical, while tax revenue and the budget balance are optimally procyclical,
implying fiscal policy countercyclicality. Then two traditional explanations of fiscal policy
procyclicality, observed in many developing countries, are modelled by Végh: imperfect
access to international capital markets (governments are cut off from international credit in
bad times, hence they are forced to cut spending and raise taxes) and weak domestic fiscal
institutions or political distortions that induce governments to undersave in good times and
oversave in bad times.
Under imperfect access to international markets, higher government spending will
have to be financed both by higher domestic and external borrowing (which drives up the
domestic interest rates above international levels) and higher tax rates. When uncertainty
about future government spending is added to incomplete capital markets, an increase in
uncertainty raises even further the procyclicality of taxes. However, under full access to
capital markets, uncertainty does not make optimal fiscal policy procyclical.
Analogously, under full access to international capital markets, a domestic political
distortion will be reflected in larger demands for government spending during good times
(when tax revue is large), which governments will not be able to resist. As a result, a second-
best optimal fiscal policy arises: in good times larger temporary tax revenue is not fully saved
(as it would be in the first-best case) but is partly spent in the form of higher government
spending and lower tax rates. Hence political-economy factors lead to procyclical fiscal policy.
2.5 A simple optimal fiscal policy rule for an open economy
We develop a simple framework for optimal fiscal policy, deriving a policy function for
government spending from stochastic government optimization subject to its budget
constraint, in a stochastic environment for an open economy characterized by key behavioral
equations.
We consider a government that pursues three fiscal policy objectives at the same time:
macroeconomic stability, expenditure stability, and debt sustainability. It does so by optimally
setting its one and only policy instrument: government expenditure.
The government objective function is an expected loss function in the quadratic
deviations of three arguments: the output gap (defined by the ratio of current to trend
9
output), the deviation of the ratio of current government expenditure to current output from
the ratio of the expenditure target to trend output, and the deviation of the ratio of current
debt to current output from the ratio of the debt target to trend output (notation is defined
below):
(1) 𝐸𝑡 [∑ 𝛽𝑠 {𝛼 (𝑌𝑠
𝑌𝑠∗)
2+ 𝛿 (
𝐺𝑠
𝑌𝑠−
�̅�𝑠
𝑌𝑠∗)
2
+ 𝜔 (𝐷𝑠
𝑌𝑠−
�̅�𝑠
𝑌𝑠∗)
2
}∞𝑠=𝑡 ]
Note that the government’s objective function embodies the short-term objective of
macroeconomic stability (the first argument) and the longer-term objectives of policy
sustainability and government solvency, reflected both in its preference of setting its
instrument (expenditure) close to its quantitative targets for spending and debt (the second
and third arguments). No political-economy arguments are included that could bias policy
toward fiscal procyclicality.
The government budget constraint is:
(2) 𝐷𝑠+1 − 𝐷𝑠 = 𝐺𝑠 − 𝑇𝐴𝑠 + 𝑟𝑠𝐷𝑠
Tax revenue is linked to current output and subject to a stochastic i.i.d. shock:
(3) 𝑇𝐴𝑠 = 𝑘𝑌𝑠 + 휀1𝑠
Output is positively affected by government spending and negatively affected by the
real interest rate, and is subject to an i.i.d. shock. The aggregate demand equation is written in
terms of the output gap and the ratio of current to target government expenditure:
(4) (𝑌𝑠
𝑌𝑠∗) = 𝛾 (
𝐺𝑠
�̅�𝑠) − 𝜂𝑟𝑠 + 휀2𝑠
Trend output grows at a constant exogenous rate but is also subject to a stochastic iid
shock:
(5) 𝑌𝑠+1∗ = (1 + 𝜌)𝑌𝑠
∗ + 휀3𝑠
The domestic interest rate deviates from the international rate proportional to the
deviation of the ratio of current debt to current income from the ratio of target debt from
trend income:
(6) 𝑟𝑠 − 𝑟𝑠∗ = 𝜙 (
𝐷𝑠
𝑌𝑠−
�̅�𝑠
𝑌𝑠∗)
The government defines the following expenditure target:
(7) �̅�𝑠 = 𝜇𝑌𝑠∗
and the following debt target:
(8) �̅�𝑠 = 𝜐𝑌𝑠∗
10
where 𝑌𝑠 is output, 𝑌𝑠∗ is trend output, 𝐷𝑠 is public debt, �̅�𝑠 is target debt, 𝐺𝑠 is government
expenditure, �̅�𝑠 is target expenditure, 𝑇𝐴𝑠 is tax revenue, 𝑟𝑠 is real interest rate. 휀1𝑠, 휀2𝑠 y 휀3𝑠
are iid stochastic tax revenue, output, and trend output shocks, respectively. Model
parameters are 𝛽, 𝛿, 𝜔, 𝑘, 𝛾, 𝜂, 𝜌, 𝜙, 𝜇, 𝜐.
Minimizing the government loss function (1) subject to equations (2)-(8) yields a
government expenditure rule of the general form:
(9) 𝐺𝑠 = 𝐺(휀1𝑠, 휀2𝑠 , 휀3𝑠, … )
where the signs of the partial derivatives of G with respect to the three stochastic shocks are
ambiguous in general. In particular, an aggregate demand shock 휀2𝑠 has an ambiguous effect
on government spending. The reason is that the government faces a trade-off in setting
spending because its loss function is complex. It depends on three different arguments, which
are affected in opposite ways – directly and indirectly – by government expenditure.
Now let’s restrict the general model to one where the interest rate is exogenous and
constant and the second and third terms of the government’s loss function are shut off,
implying:
(10) 𝛿 = 𝜔 = 𝜂 = 𝜙 = 0
The solution of this restricted government optimization yields the following
government expenditure rule:
(11) 𝐺𝑡
𝑌𝑡=
𝜇
𝛾
1
(𝑌𝑡𝑌𝑡
∗)(
𝐸𝑡(𝑌𝑡+1𝑌𝑡+1
∗ )
1+𝜌+𝜀3𝑡
(1+𝜌)𝑌𝑡−1∗ +𝜀3𝑡−1
) −𝜇
𝛾휀2𝑡
This restricted spending rule reflects unambiguous counter-cyclical behavior, as
spending is now a negative function of the aggregate demand shock.
We conclude that when the government is limited by using only one instrument to
pursue several policy objectives at the same time – like the three objectives embedded in
equation (1) – the corresponding policy trade-offs can lead to pro-cyclical fiscal policy.
However, when the use of the policy instrument is geared only to macroeconomic
stabilization, optimal policy is countercyclical, reflected in countercyclical government
expenditure.
2.6 Relations between types of fiscal target rules and fiscal performance measures
Now we turn to the expected relations between particular fiscal target rules and policy
objectives, proxied by measures of fiscal performance. We identify the three main fiscal policy
objectives and their corresponding measures of fiscal performance as columns in Table 1.
Then we identify four family of fiscal target rules and particular target rules in each family as
rows in the table. Table cells identify expected sign of the effect of a particular target rule on
the corresponding performance measure. For example, an annual budget balance rule that is
enforced every year independently of cyclical conditions increases cyclicality of government
11
spending (+). In contrast to the latter, a structural budget balance rule that is set as an average
numerical target over the business cycle, but is allowed to vary pro-cyclically, induces
counter-cyclicality of government spending (–).
How do we obtain the sign summary in Table 1? We simply use the government
budget constraint (equation (2) above), complemented by the assumption that tax revenue is
closely linked to output (automatic stabilizer) while expenditure is largely discretionary.
Hence we exclude other government objectives and restrictions by market failures and
political constraints, taken up in the literature on behavioral fiscal rules. In addition we
assume that quantitative target rules are binding, i.e., there is no fiscal space.
Table 1 Relation between Fiscal Rules and Policy Objectives
(A)
Macroeconomic
Stabilization:
Cyclicality of
government
spending
(correlations with
GDP)
(B)
Fiscal
sustainability
and solvency:
Government
deficit and debt
levels (ratios to
GDP)
(C)
Size of
Government:
Government
expenditure and
revenue levels
(ratios to GDP)
1. Budget Balance Rules (–)
Current BBR (annual) (+) (–)
Current BBR (average over the cycle) (+), (0) or (–)
Structural BBR (annual) 0
Structural BBR (average over the
cycle) (–)
2. Debt Rules (–)
Current DR (annual) (+)
Current DR (average over the cycle) (+), (0) or (–)
3. Expenditure Rules (–) (–)
Current ER (annual) (0)
Current ER (average over the cycle) (–)
4. Revenue Rules (–) (–)
Current RR (annual)
Current RR (average over the cycle)
Note: column 1 refers to cyclicality of government spending for illustrative purpose. Alternatively, the
policy objective of macroeconomic stabilization can be reflected by the cyclicality of tax rates or the
government budget.
12
The summary description in Table 1 suggest the following general results. First, for
macroeconomic stability (column 1), what matters is the particular, not the general rule.
Second, budget balance or debt targets that are enforced on a yearly basis induce pro-
cyclicality of spending, while structural or over-the-cycle rules allow for counter-cyclicality.
This ambiguous effect a fiscal on the cyclical behavior of government spending, depending on
the particular form the rule is defined, complements the previously discussed analytical
finding that the cyclical nature of a fiscal rule depends on the existence of capital-market and
political-economy distortions (Végh 2013). This ambiguity is not found in the case of the links
between rules and debt or deficit levels (for the policy objective of fiscal sustainability and
solvency, column 2) and expenditure or tax levels (size of government, column 3), assuming
that the corresponding numerical targets are well defined for achieving fiscal sustainability
and limiting government size.
3 Previous Empirical Evidence on Fiscal Rules and Fiscal
Performance
We review selectively the empirical evidence on the determinants of fiscal
performance and on the contribution of fiscal rules to fiscal performance, as well as the
reverse causality from performance to rules.
3.1 Fiscal Performance
We focus selectively on the empirical evidence regarding the role of different groups
of determinants of fiscal policy performance. Most of this evidence is based on econometric
estimations for reduced-form equations for two sets of fiscal performance measures (debt and
deficit ratios to GDP, and cyclical correlations between fiscal variables and GDP), performed
on cross-country data. The sets of determinants of fiscal performance variables include
political-economy variables, political institutions, fiscal institutions, macroeconomic variables
(business-cycle measures, trade openness, financial integration, growth, per capita GDP),
among others.
Government debt and deficits
Nerlich and Reuter (2015) conduct a comprehensive empirical analysis of the
contribution of fiscal rules and fiscal space to fiscal performance in EU countries, controlling
for a significant number of co-determinants. The government primary balance ratio to GDP is
reported to increase with financial openness, trade openness, and future age dependency, and
to decline with growth and government fragmentation. The government debt ratio to GDP
increases with future age dependency and government fragmentation, and declines with trade
openness.
For a world sample of low-income countries, Dabla-Norris et al. (2010) report that
their measures of budget institutions development (both the aggregate budget institutions
index and several of its components) raise the government primary balance and lower official
13
public external debt ratios to GDP. Macroeconomic control variables do not robustly
contribute to fiscal performance.
Eyraud and Lusinyan (2013) provide evidence on the negative contribution of vertical
fiscal integration (defined as the share of subnational net borrowing and transfers received
from general government) to the general government primary balance ratio to GDP in OECD
countries. This means that less responsibility of subnational governments for self-financing
their expenditure leads to weaker fiscal performance. The output gap (a control variable)
raises the government primary balance ratio, which suggests fiscal policy countercyclicality,
possibly due to automatic stabilizers.
Eslava (2010) presents results for government deficits in the LAC region (without
controlling for macroeconomic factors), showing that deficit ratios to GDPs are higher under
parliamentary systems and are lowered by higher levels of fiscal accountability.
Escolano et al. (2012) analyze fiscal performance (measured mostly by the primary
general government cyclically-adjusted balance ratio to GDP) in EU member countries. The
government balance is raised by the level of government debt, is lowered by the output gap
(reflecting procyclical fiscal policy), and is raised by spending decentralization to subnational
government levels.
Cyclicality of fiscal policy
Fiscal policy cyclicality is measured in empirical work as a rolling correlation
coefficient or a regression coefficient between a fiscal performance aggregate (in level or in its
cyclical component) and GDP (in level or its cyclical component).
Végh (2013, pp. 484-6) reports for a world sample cross-country correlations
between the cyclical components of government spending and GDP, and between the cyclical
components of tax rates and GDP, suggest that fiscal policy tends to be countercyclical in
industrial economies and procyclical in emerging-economy and developing countries Further
evidence suggests that the causality is largely from the GDP cycle to government spending and
taxes, and not the reverse (Ilzetzki and Végh 2008, Végh and Vulletin 2012).
Several explanations are offered to explain these stylized facts on fiscal policy
cyclicality. First, developing and emerging economies face more stringent capital market
imperfections. They tend to be cut off from international financial markets in bad times
(Aizenman et al., 2000) and, more generally, they face more incomplete financial markets
(Riascos and Végh 2005). Second, they have weaker fiscal institutions in place, making fiscal
policy more prone to agency and political-economy problems that imply spending more in
good times and less in bad times (Talvi and Végh 2005).
Calderón and Schmidt-Hebbel (2003) and Calderón et al. (2004) identify a causal
factor that drives the difference in fiscal policy cyclicality within a sample of developing
countries: the foreign debt premium or spread, a measure of the cost of foreign lending and of
weak policy credibility. They find that at low (high) spreads, fiscal policy, measured by the
cyclical component of government spending is countercyclical (procyclical). Related work by
Calderón et al. (2016) focuses on different measures of institutional quality and their relation
to policy cyclicality. For a world sample of 112 industrial and developing countries, they
14
report that at high (low) levels of institutional quality, fiscal policy is countercyclical
(procyclical).
Nerlich and Reuter (2015) estimate fiscal policy procyclicality as a time-varying
regression coefficient for growth in an equation for government spending that controls for
other variables. Then the authors regress the estimated procyclicality coefficients, finding that
fiscal space (the difference between the debt limit and current debt, a la Ghosh et al. 2013)
and trade openness raise procyclicality, while growth and age dependency reduce cyclicality.
Finally, Dabla Norris et al. (2010) report that their measures of budget institutions
development reduce fiscal policy cyclicality in low-income countries.
3.2 Fiscal Rules and Fiscal Performance
We start by reviewing the scant evidence on the contribution of fiscal rules to fiscal
performance. Then we turn to the reverse causality: evidence on the contribution of fiscal
performance measures to the adoption of fiscal rules.
From rules to fiscal policy cyclicality
Eslava (2010) reports that fiscal rules reduce the levels of government debt and
deficits in Latin American and Caribbean countries.
In their above mentioned work, Nerlich and Reuter (2015) report several fiscal
performance measures that are significantly affected by having fiscal rules in place,
controlling for many other determinants (several were mentioned above). First, fiscal space is
increased by having any fiscal rule in place, and in particular a budget balance and an
expenditure rule. Second, an expenditure rule raises the government primary balance and
reduces government debt. Third, discretionary current expenditure is raised and
discretionary current revenue is lowered by having any fiscal rule in place. However, this
result may be spurious as it refers to having any fiscal rule in place, because the expected sign
of the effect of any rule should depend on which rule is implemented, as discussed above.
Finally, the effects of fiscal rules on procyclicality are reported to be positive for a budget
balance rule and negative for an expenditure rule, but the magnitude of the latter effects is
diminished by the size of fiscal space, a variable that interacts with fiscal rules. Also in this
case the results do not seem to be fully consistent with the ambiguity of the effects of general
rules on policy cyclicality, as extensively discussed above.
Reverse Causality: from Fiscal Performance to Rules
An issue that is of significance for our estimations below on the effects of fiscal rules
on different measures of fiscal performance is potential reverse-causality running from fiscal
performance to the implementation of rules. In fact, in previous work, we have reported
significant contributions of two fiscal performance measures (government budget balance
and procyclicality of government expenditure) to the likelihood of having a national fiscal rule
in place, controlling for a host of other macroeconomic, structural, institutional, and political
conditions. These results underscore the importance of instrumenting properly fiscal rules in
our econometric analysis, as reported below.
15
4 Are Small Economies Different?
There is an extensive literature documenting the notion that small states share
characteristics that are absent in larger economies (Favaro, 2008). Most of the world’s small
states are very young, having achieved independence in the past 50 years. Political
independence meant that services formerly provided by colonial institutions had to be
supplied by fledgling national institutions. Furthermore, there have always been questions
about the implications of a small domestic market for small states’ income levels and growth
potential. A small domestic market limits capacity to exploit economies of scale and diversify
risk. Other features of small states are that they receive more foreign direct investment and
worker remittances, and tend to have relatively larger governments with higher debt and
deficit levels.
Yet small states as a group tend to grow as fast as other emerging economies.
Compensating factors for their size limitation include their reliance on and proximity to world
markets (despite their larger export concentration), their relatively higher endowments of
human and physical resources; and the policies, institutions, and regulations that many small
states have adopted to facilitate their integration into world markets.
Moreover, Easterly and Kraay (2000) conclude that, controlling for location, small
states have higher per capita GDP levels than other states. This income advantage is largely
due to a productivity advantage, which is evidence against the notion that small states suffer
from an inability to exploit returns to scale. While small states have similar per-capita growth
rates than other economies, they suffer from higher instability, which is in part due to their
larger terms-of-trade volatility, due to their larger openness.
Based on current data, Table 2 confirms several features small states that were
identified in previous papers. We split countries along the 1.5 million inhabitants threshold
level (the World Bank’s limit of a small state) and conduct simple mean tests to identify the
main differences between large and small economies. Small countries have higher levels of
per capita GDP and enjoy higher government stability, while their political checks and
balances and political participation level (a measure of democracy) are slightly less
developed. Small states are significantly more open to foreign trade, while their export base
tends to be less diversified, although they enjoy lower natural resource rents. Small
economies display a degree of financial openness that is similar to other countries are
similarly open regarding capital-market integration equally open in terms of financial
markets. They are more likely to adopt more rigid exchange-rate regimes despite receiving
much larger inflows of worker remittances. Small and large states share similar levels of tax
revenue instability and inflation levels, and there is no evidence that business cycles are more
pronounced in smaller economies.
16
Table 2 Stylized facts of Small versus Large Economies
Variable
Large Economies
(more than 1.5
million inhabitants)
Small States
(less than 1.5 million
inhabitants)
Mean Test
large–small
Mean Std. Dev Mean Std. Dev
Business Cycles (index) 0 0.03 0 0.03 -0.14
Checks and Balances (index) 2.76 1.77 2.38 1.35 7.09*
Dependency Ratio (% population) 60 0.12 60 0.12 -0.04
Development Level (GDP per capita) 2,582 1.65 4,891 1.4 -12.18*
Exports Concentration (index) 2.49 0.61 3.30 0.55 -13.46*
Financial Development (% of GDP) 40% 0.88 37% 0.94 2.55*
Financial Openness (index) 0.47 0.37 0.44 0.35 2.16*
Fiscal Revenue Instability (index) 0.07 0.09 0.07 0.08 -1.00
Fixed Exchange Rate (index) 0.29 0.45 0.62 0.49 -19.54*
Government Stability (index) 2.0 2.11 3.5 2.3 -4.58*
Political Participation (index) 5.05 4.07 4.17 4.29 4.81*
Price Instability (annual inflation) 9.9 0.15 5.7 0.08 12.24*
Resource Rents (US$ per capita) 7,476 1.9 5,249 2.86 3.43*
Sovereign Wealth Funds (index) 0.05 0.12 0.05 0.15 0.60
Trade Openness (% of GDP) 64 0.55 107 0.41 -32.74*
Workers Remittances (% of GDP) 0.83 1.96 1.34 1.86 -6.10*
(*) significant at 95% confidence or higher.
Source: own elaboration based on data from World Bank Database.
The fact that small states exhibit structural differences compared to larger economies
suggests that fiscal rules might affect small economies differently. It also suggests the need to
control for economic size in our subsequent statistical analysis. While we use an arbitrary
threshold level to identify small states – a population of 1.5 million, as noted above – we
obtain similar econometric results when employing a continuous variable for population size.
5 Empirical Methodology
Empirical assessment of the impact of rules on fiscal outcomes (balance, debt, and
procyclicality) has to address three relevant issues: (a) fiscal rules tend to display high
persistence, thereby requiring the use of dynamic models, (b) countries are highly
idiosyncratic regarding unobservable features in their budget structures, indicating the
existence of individual country effects, and (c) reverse causation and endogeneity are very
likely, since governments may decide to implement rules as a result of their fiscal stance.
Standard dynamic panel data models of the Arellano-Bond family can be used to address the
first two issues but their results are hampered by the third. Therefore we complement
17
dynamic panel data techniques with a novel approach to building a consistent instrument for
the choice of implementing a fiscal rule.
From a methodological viewpoint, unveiling the impact of rules on fiscal performance
can be casted as the study of whether a treatment (implementing a fiscal rule) has any
discernible effect on fiscal performance (FP) of country “i” at time “t”, which we denote
generically by 𝐹𝑃𝑖𝑡. A general model would be of the form:
(12) 𝐹𝑃𝑖𝑡 = 𝑓(𝛼𝑖 , 𝑥𝑖𝑡 , 𝐷𝑖𝑡 , 𝐹𝑃𝑖𝑡−1)
where parameter 𝛼𝑖 reflects cross sectional heterogeneity (i.e., individual effects), 𝑥𝑖𝑡
represents the set of fundamentals – other than a fiscal rule, FR – that determine fiscal
outcomes, and 𝐷𝑖𝑡 denotes the implementation of a FR. The presence of 𝐹𝑃𝑖𝑡−1 indicates the
dynamic nature of our models, which captures the inertia of fiscal performance variables.
Identifying causal effects from observational data is often quite difficult as it depends
crucially on the validity of the implicit or explicit identification assumptions that are
particular to the corresponding empirical approach. As noted by Jordá and Taylor (2016), the
divergence of results in fiscal policy analysis can be the result of poor identification conditions
and, indeed, much of the variation in results is likely to be the consequence of endogenous
factors being either ignored or inadequately treated by the econometrician. In the context of
panel data, this would indicate the need of properly specifying the determinants of fiscal
performance, as well as controlling for potential endogeneity of the treatments. Furthermore,
modelling treatment is not as straightforward as it might seem, since it requires separating
the determinants of fiscal outcomes from those that led to the enactment of such rules. In
what follows we first address the issue of the modelling strategy for the treatments (i.e., the
implementation of the fiscal rule) and later the issue of potential endogeneity of fiscal rules
and other controls that co-determine fiscal outcomes.
5.1 Fiscal Rules as Treatments
In principle, the analysis of the effects of policy variables on outcomes of interest
ought to consider the intensity of such policies by using continuous variables. The imposition
of a fiscal rule as well as any other fiscal institutions (e.g., fiscal councils, sovereign wealth
funds), on the contrary, can be viewed as a once-and-for-all event (under the assumption that
such rule does not change in time). In this context it seems appropriate to use a dummy
variable (𝐷𝑖𝑡) taking value 1 whenever the rule is in place and zero otherwise. We further
assume that such policy is determined by 𝐷𝑖𝑡 = 𝐷(𝑤𝑖𝑡 , 𝜓𝑖, 휀𝑖𝑡) where 𝑤𝑖𝑡 is the conditioning
set (historical data), 𝜓𝑖 refers to the parameters of the implied policy function in country i and
휀𝑖𝑡 is an idiosyncratic source of random variation. Historical data would include past
observations on fiscal performance and other variables that may inform about the presence of
the rule. Therefore, 𝐷(𝑤𝑖𝑡 , 𝜓𝑖, . ) refers to the systematic component of policy determination.
18
Following Angrist et al. (2013) we define the potential outcomes of a policy as given
by 𝐹𝑃𝑖,𝑡+ℎ(𝐷𝑖𝑡) − 𝐹𝑃𝑖,𝑡, i.e., the value that the observed performance variable 𝐹𝑃𝑖,𝑡+ℎ − 𝐹𝑃𝑖,𝑡
would take when the rule is not enacted (𝐷𝑖𝑡 = 0) and when it is (𝐷𝑖𝑡 = 1). Therefore the
difference 𝐹𝑃𝑖,𝑡+ℎ − 𝐹𝑃𝑖,𝑡 refers to the cumulative change in outcome from 𝑡 to 𝑡 + ℎ.
Considering that the impact of a fiscal rule on the government stance might take time to yield
its potential effects, we use five-year averages in our subsequent econometric analysis.
Consider now expressing equation (12) in linear form:
(13) 𝐹𝑃𝑖𝑡 = 𝛼𝑖 + 𝛽𝑖𝑥𝑖𝑡 + 𝜃𝐷𝑖𝑡 + 𝛾𝐹𝑃𝑖𝑡−1 + 𝑢𝑖𝑡
Standard panel data estimators, such as pooled or fixed-effects models, are
inappropriate to parameterize equation (13) because there would be correlation between the
individual effect and the error term. The pooled estimator, which ignores heterogeneity
altogether, induces correlation between the residual and the lagged endogenous variable. For
the fixed effects estimators, on the other hand, Nickell (1981) shows that the demeaning
process which subtracts the individual’s mean value of the variables creates a correlation
between regressors and the error term. The resulting correlation biases the estimate of the
coefficient of the lagged dependent variable which is not mitigated by increasing the number
of individual units.
One solution to this problem involves taking first differences (∆) of the original model.
(14) ∆𝐹𝑃𝑖𝑡 = 𝛽𝑖∆𝑥𝑖𝑡 + 𝜃∆𝐷𝑖𝑡 + 𝛾∆𝐹𝑃𝑖𝑡−1 + ∆𝑢𝑖𝑡
The first difference transformation removes the individual effect (and any other
constant) but it comes at the cost of inducing correlation between the differenced lagged
dependent variable and the disturbance process (the former contains 𝐹𝑃𝑖𝑡−1 and the error
term contains 𝑢𝑖𝑡−1). But with the individual fixed effects swept out, instrumental variables
estimators are available. As shown by Anderson and Hsiao (1981), we may construct
instruments for the lagged dependent variable from the second and third lags of 𝐹𝑃𝑖𝑡, either in
the form of differences or lagged levels. As argued by Blundell and Bond (1998) and others,
the instrumental variables approach does not exploit all of the information available in the
sample. By doing so in a generalized method of moments (GMM) context, we may construct
more efficient estimates of the dynamic panel data model. Monte Carlo studies have shown
that estimated asymptotic standard errors of the efficient two-step GMM estimator can be
severely biased downward in small samples. Therefore we, use Windmeijer’s (2005) finite
sample correction in order to obtain robust estimators.
Beyond the unbiased estimation of the parameters of the model, appropriate
estimation of 𝜃 requires 𝐹𝑃𝑖,𝑡+ℎ(𝐷𝑖𝑡) − 𝐹𝑃𝑖,𝑡 to be orthogonal to 𝐷𝑖𝑡|𝑤𝑖𝑡 (for all h). This
conditional independence assumption plays an important role and allows us to identify the
average causal effect of a policy intervention relative to a baseline on the outcome variable at
time t+h using local projections (Jordá and Taylor, 2016):
19
(15)
𝐸[(𝐹𝑃𝑖,𝑡+ℎ(𝐷𝑖𝑡 = 1) − 𝐹𝑃𝑖,𝑡) − (𝐹𝑃𝑖,𝑡+ℎ(𝐷𝑖𝑡 = 0) − 𝐹𝑃𝑖,𝑡)] =
𝐸[𝐸(𝐹𝑃𝑖,𝑡+ℎ − 𝑦𝑖,𝑡|𝐷𝑖𝑡 = 1; 𝑥𝑖,𝑡) − 𝐸(𝐹𝑃𝑖,𝑡+ℎ − 𝐹𝑃𝑖,𝑡|𝐷𝑖𝑡 = 0; 𝑥𝑖,𝑡)] =
𝜃ℎ[(𝐷𝑖𝑡 = 1 − 𝐷𝑖𝑡 = 0)]
If the conditional independence assumption fails, OLS applied to equation (15) will
deliver a biased and inconsistent estimate of 𝜃. This is a case known as the endogenous dummy
variable problem (Heckman, 1978).5 This is most likely the case in the context of fiscal rules
since they are imposed precisely as the result of the perceived need of enacting additional
mechanism to achieve or secure policy outcomes. Instrumental variables can be brought in to
fix this inconsistency, but need to meet two well-known conditions. First, they need to be
independent of the unobserved selection mechanism. Second, the instruments 𝑧𝑖𝑡 need to be
predictive for 𝐷𝑖𝑡. Assuming these two conditions are met, estimation of the response to policy
interventions using local projections will deliver a consistent estimate of 𝜃.
Estimation of these conditional expectations can be simplified considerably when a
model for the policy variable 𝐷𝑖𝑡 is available. Angrist and Kuersteiner (2011) refer to the
predicted value from such a policy model the “policy propensity score”. The policy propensity
score acts as a dimension-reduction device and is meant to ensure the estimation of the policy
response (the average treatment effect in microeconomics parlance) is consistent under the
main assumption.
5.2 On the Endogeneity of Fiscal Rules
A legitimate concern is whether policy instruments, such as fiscal rules, are truly
exogenous in models of fiscal performance. While one would expect that the adoption of new
rules and institutions would induce better fiscal outcomes, it is nevertheless possible that said
policy instruments are adopted after an economy has already improved its fiscal indicators or
is firmly on a path of consolidating them. If there is reverse causality, the estimation of the
impact of fiscal rules on performance will be biased (the estimator is inconsistent).
This problem of reverse causality is an important issue that cannot be adequately
addressed by the standard time-series practice of using lagged values and GMM conditions for
the potentially endogenous regressor (such as Arellano and Bond, 1991). We deal with the
possible endogeneity of fiscal rules using instrumental variables. Assuming that we have a set
of valid instruments 𝑧𝑖𝑡 for adopting a fiscal rule (not including the elements that determine
fiscal performance in 𝑥𝑖𝑡), we can consistently estimate our models by the following three-
stage procedure:
5 Assume for simplicity that all projection slopes but 𝜃 are zero. Then, if 𝑐𝑜𝑣(𝐷𝑖𝑡 , 𝑢𝑖𝑡) ≠ 0, the estimator
of interest will be: �̂� = 𝜃 +𝑐𝑜𝑣(𝐷𝑖𝑡,𝑢𝑖𝑡)
𝑉𝑎𝑟(𝐷𝑖𝑡). The direction of OLS inconsistency is determined by the sign of
𝑐𝑜𝑣(𝐷𝑖𝑡 , 𝑢𝑖𝑡). For example, if countries are more likely to have a fiscal rule when performance is good, we have that 𝑐𝑜𝑣(𝐷𝑖𝑡 , 𝑢𝑖𝑡) > 0 and the OLS estimator will over-estimate the true value (asymptotically).
20
(1) In the first stage, we estimate by maximum likelihood a probit model of the
determinants of having a fiscal rule in place and compute the predicted probability FR̂.
The dependent variable is binary, taking value 1 if the fiscal rule is in place and value 0
otherwise. The predicted value of this model is a valid, yet inefficient, instrument for
the fiscal rule.
(2) In the second stage, we regress the fiscal rule on FR̂ and 𝑥𝑖𝑡 using OLS and compute
the fitted values FR̿̿̿̿ . This step ensures the conditional independence assumption that
is crucial to identify in a causal manner the role of rules on fiscal performance (Adams
et al., 2009, Wooldridge, 2002, p. 623).
(3) In the third stage, we regress a fiscal performance variable, 𝐹𝑃𝑖𝑡, on 𝑥𝑖𝑡 and the fitted
values of the second stage, FR̿̿̿̿ . In this stage we control for a number of independent
variables – including institutions, exchange rate and monetary regimes, financial
development, political variables and overall economic development – in order to
isolate the true contribution of rules to the four different measures of fiscal
performance.
This procedure is different from the “pseudo-IV” procedure of running an OLS
regression of 𝐹𝑃𝑖𝑡 on FR̂ and 𝑥𝑖𝑡. In the latter case, consistency is not guaranteed unless the
first stage is correctly specified, and the standard errors need to be adjusted. There are many
advantages of our approach. First, it takes the binary nature of the endogenous variable into
account. Although the two-stage least squares consistency of the second stage does not hinge
on getting the functional form right in the first stage (see Angrist and Krueger, 2001), two-
stage least squares leads to biased estimates in finite samples and it is not known how
misspecification in the first stage may affect this bias. Second, unlike some of the alternative
procedures, it does not require the binary response model of the first stage to be correctly
specified. Third, although some regressors are generated in the first stage, the standard IV
standard errors are still asymptotically valid (Wooldridge 2002).
6 Empirical Analysis I: Instrumenting Fiscal Rules
In the first part of our empirical analysis we generate an instrument for the presence
of fiscal rules (as well as for the other potentially endogenous fiscal institution, sovereign
wealth funds), which we then use to obtain an unbiased estimate the effect of rules on
outcomes. We use annual data for the period 1985-2015. Macroeconomic data tend to be very
and unreliable before 1985, especially for fiscal variables in developing and emerging-market
economies, as a result of changes in definitions and measurement methods, structural
changes, and lack of data. Our sample is restricted to 120 countries but they represent all
world regions (Appendix Table 1 lists sample countries). The total number of observations is
around 2,750, depending on data availability.
21
In the first stage of our methodology, we estimate a probit model for the incidence of
FRs, defined as a binary dummy variable taking value 1 if there is a fiscal rule in place and 0
otherwise:
(16) 𝑃𝑟𝑜𝑏(𝐹𝑅𝑖𝑡 = 1|𝑥𝑖𝑡 , 𝑤𝑖𝑡) = 𝜙(𝜋𝑖 + 𝜌𝑥𝑖𝑡 + 𝑤𝑖𝑡)
where ϕ(·) is the cumulative distribution function for a standardized normal random variable,
𝑤𝑖𝑡 is the vector of instruments, and 𝑥𝑖𝑡 is a vector of other control variables. It is important to
highlight that our IV approach does not require this specification to be correct nor the
estimation to be efficient. It only requires the instruments to be correlated with the
probability of having a fiscal rule in place.
For independent variables in equation (16), the literature suggests several potential
determinants of the likelihood of implementing a fiscal rule. Following the results in Schmidt-
Hebbel and Soto (2017), we group these potential determinants of fiscal rules in five areas:
Institutional Variables
Fiscal rules are likely to be the outcome of particular political regimes and institutions.
By constraining policymakers in the design and execution of the budget, in a way that is
relatively transparent and subject to open monitoring, fiscal rules reflect more transparency,
stronger democratic accountability, less discretion, and less corruption. Therefore, our first
political determinant is a standard measure of political participation or democracy (see
Appendix Table 2 for the sources and definitions), for which we expect a positive sign.
At the constitutional level, the distinction between federal and unitary government is
likely to make a difference in the adoption of fiscal rules and, more importantly, in the type of
rule selected. In federal countries, the fiscal sovereignty of federal governments is weaker
than that enjoyed by central governments in unitary countries. The vast literature on fiscal
federalism attests to the important differences in the conduct and outcome of fiscal policy
between federal and unitary countries (e.g. Feld and Schnellenbach, 2010). We expect federal
governments to be more likely to adopt fiscal rules –in particular, a budget balance rule—than
unitary governments, because they strengthen their bargaining position with respect to the
federated states or provinces. Likewise, federal economies ought to be less likely to adopt a
revenue rule over which the central government has little control. For this reason, we include
a binary dummy variable taking value 1 if there is a federal government and 0 otherwise,
which we label as federalism.
There is evidence suggesting that rules reflect an implicit contract between
governments and voters. In other words, they signal a government commitment to maintain
mutually agreed standards of fiscal discipline (Debrun and Kumar, 2007). Therefore, we
include a measure of political checks and balances.
As the political instability of governments makes it difficult to pre-commit to rules,
FRs are more likely to be adopted and continued over time under conditions of government
22
stability. Hence we include a government stability measure as a potential regressor and expect
a positive effect on the existence of a fiscal rule. We use the International Country Risk Guide
(ICRG) measure which defines government stability as a combination of a government’s
ability to carry out its declared program and its ability to stay in office.
Exchange Rate and Monetary Regimes
Two types of exchange rate regime are likely to affect the adoption of fiscal rules. A
super-hard exchange rate regime in the form of absence of a national currency due to the use
of an international currency as a result of monetary union membership is likely to raise the
likelihood of adoption of supranational fiscal rules that are then implemented at the national
level. This is to reduce the incidence of moral hazard in member countries’ conduct of
sovereign fiscal policy. Therefore, we include monetary union membership as a regressor.
Regarding the connection of fiscal rules and exchange regimes, the literature has
explored mainly the reversed causality, i.e., stemming from fiscal discipline to exchange
regime stability. The conventional view is that pegs provide more fiscal discipline than floats
(Giavazzi and Pagano, 1988 and Frenkel et al., 1991). If governments adopt a lax fiscal policy
under a fixed exchange rate, this would lead to a speculative attack on reserves and,
consequently, result in currency devaluation and a large political cost that the policymaker
would prefer to avoid by imposing discipline on fiscal outcomes. Alternatively, Tornell and
Velasco (2000) stress that, under reasonable conditions (linked to governmental uncertainty
about re-election and lack of access to capital markets), more fiscal discipline is attained
under floats, where fiscal mismanagement leads to devaluation and inflation in the short term.
Under pegs, unsustainable fiscal policy leads to higher debt and lower reserves in the short
term, postponing the costs of devaluation and inflation to the future. Hence we include as a
second exchange rate regime measure a binary variable for a fixed exchange rate regime. We
use a de jure measure taking value 1 if the country has a fixed exchange system and zero
otherwise. Fixed exchange systems include dollarization, currency boards, and monetary
unions. Data was taken from Reinhart and Rogoff (2004) and extended to 2015 using IMF
country reports.6 Considering the preceding arguments, the expect effect of a hard peg of the
currency on the likelihood of having a fiscal rule in place is ambiguous.
Turning now to monetary policy, there is significant theoretical and policy consensus
that the absence of fiscal dominance is a pre-condition for the success of inflation targeting
(see Minea and Villieu, 2009). Fiscal dominance –the need to rely on central bank resources,
ultimately seigniorage– is less likely when a government commits to a fiscal rule. Lucotte
(2012) finds that inflation targeting has a significant positive effect on public revenue
collection in thirteen emerging economies. Hence we include a discrete variable for countries
in which monetary policy is based on an inflation-targeting regime (inflation target). We
expect that an inflation-targeting regime increases the likelihood of having a fiscal rule in
place.
6 We also tested our models using a de facto measure of exchange regime based on observed movements in the nominal exchange rate following Shambaugh (2015) classification. The results are qualitatively similar but the sample is shorter so that we prefer the de jure measure.
23
Capital Account Openness and Financial Development
Financial development could have a positive influence on the likelihood of having
fiscal rules in place through two channels. First, both domestic financial development and
stronger integration into world capital markets increase governments’ access to domestic and
external debt financing and subject governments to closer scrutiny of fiscal sustainability on
the part of financial market analysts and rating agencies. This strengthens the case for
adopting fiscal rules that commit governments to fiscal prudence and solvency. Second, if
domestic financial markets are deeper, and integration into world capital markets is full and
comprehensive, governments will be more likely to access domestic or external funding
during cyclical downturns. This reinforces governmental adoption of fiscal rules that
minimize fiscal pro-cyclicality or strengthen fiscal counter-cyclicality. Therefore, we include
one variable that reflects domestic financial development and another that measures
international financial integration or financial openness as potential determinants of having
fiscal rules in place.
Overall Economic Development
We use real GDP per capita to control for the overall development level. Some studies
focus on the reverse causality, i.e. on the impact of fiscal rules on economic growth (e.g.
Castro, 2011). Here we focus on the causality from the level of development to the likelihood
of having a fiscal rule in place. This hypothesis embodies the stylized fact that governments in
richer economies have more human and financial resources available to undertake the
complex tasks of adopting, complying with, monitoring, and evaluating the operation of a
fiscal rule.
Next we include the population dependency ratio, i.e. the ratio of the population under
15 and over 64 years of age to those falling within the age range 15–64. As the ratio rises, so
does the expected demand for higher government spending on social programs in support of
the young and the elderly (for childcare, education, pensions, and health). This makes it more
attractive for governments to commit to a fiscal rule for prudential reasons (e.g., to avoid
accumulating unsustainable government debts) thereby increasing the likelihood of putting
one in place.
External Fiscal Shocks
One main role of fiscal rules is to smooth out the effect of transient macroeconomic
phenomena from government budgets. We control for the two main external shocks that have
been documented to have significant impact on fiscal variables: international workers’
remittances and natural resource rents. On one hand, evidence indicates that remittances
appear to be strongly procyclical vis-à-vis sending country income and tend to be spent on
consumption of both imported and domestically produced goods, rather than on investment
(Abdih et al., 2012). Therefore, shocks in sending countries are transmitted via remittances to
the public finances –specifically, tax revenues—of receiving countries. On the other hand, in
countries where the resource revenue constitutes a large component of total government
revenues, commodity price fluctuations will have a direct impact on public spending
24
(Bjørnland and Thorsrud, 2015). Many resource-rich countries are therefore advised to adopt
some type of fiscal policy framework (i.e., a fiscal spending rule), which, if operated
countercyclically, should shelter the economy from commodity price fluctuations prevent
over-spending on the part of the government. Consequently, we include in our models both a
measure of natural resource rents and the amount of workers remittances received by a
country (both as share of GDP).
Our econometric models update and extend previous studies by Schmidt-Hebbel and
Soto (2017), Kumar et al. (2009), and Calderón and Schmidt-Hebbel (2008a). We lag the
independent variables in order to avoid simultaneity problems. The results of the estimations
are presented in Table 3. First, note that the null hypothesis of the poolability test is strongly
rejected in all models, indicating that there exists country heterogeneity in the form of
individual effects and that a panel probit model is an adequate specification. Second, the
models present a rather high fit to the observed data, as shown in relatively high pseudo R2
indicators. This would ensure that our instruments are highly –but not perfectly—correlated
to fiscal rules. Third, note that most right-hand side variables are statistically significant and
that the evidence supports the above discussion on the determinants of having a fiscal rule in
place. Naturally, not all variables have a bearing on implementing a particular rule but, when
significant, the estimated parameters have the expected sign. Recall, nevertheless, that when
creating an instrument the only consideration is that right-hand side variables must be
correlated with fiscal rules and that neither the model must be well specified nor the
estimated parameters must have an economic interpretation.
As shown in Table 3, our results reveal that some variables strongly influence the
probability of observing a fiscal rule in place in a country, independently of the type of rule.
Economies that feature lower levels of resource rents, higher population dependency, and
larger financial openness are more likely to have implemented any of the four specific fiscal
rules we consider. Likewise, countries that conduct monetary policy using modern inflation-
targeting schemes or belong to a monetary union are also more likely to have fiscal rules in
place.
Other variables influence selectively the choice of particular fiscal rules. For example,
more developed economies are more likely to adopt expenditure and/or revenue rules, but
not budget balance rules and debt rules. Finally, some variables have opposite effects on the
selecting different rules; for example, federal constitutions raise the likelihood of adopting
budget balance rules and lower the likelihood of adopting revenue rules.
In order to save space, we present the empirical results of our second stage in
Appendix Table 3. In addition to the first-stage fundamentals, the models include the
predicted probability of having a fiscal rule in place, (𝐹�̂�). While results are in general in line
with those reported in Table 3, the estimated parameters for the different 𝐹�̂� are highly
significant, indicating the importance of this second stage for obtaining efficient instruments
of fiscal rules.
Finally, we use the same methodological approach to generate instruments for
another potentially endogenous fiscal institution: sovereign wealth funds. These are state-
25
owned investment vehicles that invest globally in various types of assets ranging from
financial to real to alternative assets. Around one half of the sovereign wealth funds in the
world have been designed and implemented for fiscal stabilization goals. We follow Elbadawi
et al. (2017) to select the appropriate set of fundamentals and generate an instrument for the
presence of external sovereign wealth funds. The detailed discussion and empirical results are
presented in Appendix 1.
Table 3 Stage 1: Probit Regression Models for Fiscal Rules
------------------------------------------------------------------------------------------------------------ National Rules --------------------------------------------------------------------------- Any Rule Any Expenditure Revenue Budget Balance Debt ------------------------------------------------------------------------------------------------------------ Political 0.157*** 0.0873 0.283** 0.707** 0.0113 -0.0390 Participation (3.15) (1.14) (2.24) (2.34) (0.15) (-0.68) Federalism -4.720* 0.780 -0.147 -3.616*** 4.051*** 1.091 (-1.85) (0.60) (-0.13) (-2.82) (3.01) (0.97) Checks and 0.141*** 0.0552 -0.117 0.0576 0.163*** 0.0142 Balances (3.36) (1.44) (-1.51) (0.39) (3.18) (0.14) Government -0.00372 0.0622* -0.00448 0.182* -0.0257 0.00262 Stability (-0.12) (1.73) (-0.11) (1.95) (-0.57) (0.05) Monetary Union 13.79*** 1.273*** 1.165*** 1.923** -3.097* -1.668** (6.23) (4.19) (3.08) (2.21) (-1.81) (-2.13) Fixed Exchange 1.102*** -0.0879 -0.477 0.636 -0.418 1.062*** Regime (3.82) (-0.36) (-1.61) (1.03) (-0.89) (2.77) Inflation Target 1.516*** 1.333*** 0.603* -0.388 1.304*** 1.773*** (5.38) (5.43) (1.89) (-0.61) (4.48) (4.87) Financial 2.334*** 1.387*** 2.486*** -1.629 1.801*** -0.198 Openness (5.98) (3.38) (4.37) (-1.35) (3.14) (-0.40) Financial -0.322* 0.421** 0.0678 1.000** 0.629*** 0.358 Development (-1.69) (2.55) (0.29) (2.14) (2.63) (1.42) Economic 1.834*** 0.559 1.284* 2.315*** 0.208 -0.934** Development (4.84) (1.49) (1.77) (3.23) (0.59) (-2.31) Dependency 22.56*** 27.94*** 17.01*** -10.99 31.08*** 31.84*** Ratio (6.38) (7.14) (4.42) (-1.29) (6.51) (7.77) Resource -4.244*** -1.376*** -0.802 -1.703*** -0.778** -0.0172 Rents (-10.74) (-2.65) (-1.53) (-3.93) (-2.06) (-0.05) Workers 0.343*** 0.0127 0.156 1.012*** 0.163 -0.181* Remittances (3.82) (0.17) (1.39) (2.61) (1.57) (-1.79) Constant -173.4*** -146.0*** -101.6*** -14.63 -150.3*** -131.0*** (-11.12) (-9.97) (-5.25) (-0.47) (-7.94) (-9.54) ------------------------------------------------------------------------------------------------------------ Poolability Test 3.971*** 3.265*** 2.606*** 3.556*** 3.492*** 3.176*** (18.16) (10.42) (4.57) (12.79) (14.08) (13.46) ------------------------------------------------------------------------------------------------------------ Observations 2744 2744 2744 2744 2744 2744 Number of Countries 118 118 118 118 118 118 Pseudo R squared 0.79 0.71 0.62 0.70 0.71 0.69 ------------------------------------------------------------------------------------------------------------ t statistics in parentheses * p<0.1, ** p<0.05, *** p<0.01
26
7 Empirical Analysis II: Assessing the Impact of Fiscal Rules
In the first part of this section we discuss in more detail the measures of fiscal
performance we use in the empirical analysis. In subsequent subsections, we estimate
econometric dynamic panel-data models to evaluate how fiscal rules affect the fiscal stance.
The regression analysis is conducted for around 120 economies over 1985-2015, using
averages of five-year periods to reduce the impact of temporary events. Each country has a
minimum of three and a maximum of seven non-overlapping five-year observations (the
panel is slightly unbalanced due to missing data). A minimum of three observations per
country is required to run the instrumental variable methodology outlined above. Since one
observation is reserved for instrumentation, the first five-year period corresponds to 1990-
1994. The total number of observations is around 430, depending on data availability.
7.1 Measures of fiscal performance
For our empirical analysis we focus on four measures of fiscal performance. Our
selection of measures is determined by data availability and our reviews carried out in
sections 2 and 3. Hence we focus our analysis on two indicators of procyclicality (government
expenditures and fiscal balances) and two of sustainability (fiscal balance and debt).
We measure fiscal procyclicality as the rolling correlation between the cyclical
components of government expenditure and of GDP, both at constant prices. We also use a
similar measure for fiscal balance procyclicality. The underlying data is from World
Development Indicators and the IMF World Economic Outlook database for general
government expenditure, fiscal balance, and GDP. To estimate the cyclical components, we de-
trend all variables using a bandpass filter (Hodrick and Prescott, 1997) and then compute ten-
year rolling correlations of the cyclical components of government consumption and GDP. We
also use alternative windows and periods for computing the rolling correlations (3 and 5
years) to check for robustness.
We measure the government budget balance as a ratio to GDP on annual basis. For
government balances some countries report data for the central government while others
include subnational government units and report data for the general government. We use the
IMF Fiscal Database as the main source for our measure of the fiscal balance of the central
government, supplemented, when necessary, by country data from ECLAC, the African
Development Bank, and the Asian Development Bank.
For government debt we use the IMF’s measure of “General government gross debt”,
comprising all liabilities requiring future payments of interest and/or principal. This measure
(expressed as ratio to GDP) includes debt liabilities in the form of SDRs, currency and
deposits, debt securities, loans, insurance, pensions and standardized guarantee schemes, and
other accounts payable.
27
7.2 Determinants of Fiscal Performance
In order to isolate the contribution of fiscal rules to fiscal performance, we control for
other determinants of performance, as suggested by theory and previous evidence, reviewed
in section 3. Our control variables include other fiscal institutions, political institutions, the
level of development, macroeconomic regimes, and macroeconomic controls (including
cyclical conditions). The fiscal implications of many of these variables were discussed above;
here we keep their description to a minimum.
Access to borrowing by governments
Capital account openness and financial development largely determine the ability of
governments to borrow money in domestic and external markets to fulfill their financial
needs. Measuring access to foreign borrowing is not an easy task but accessing funds and
covering financial needs should be easier in economies with more open capital accounts. We
use the de jure measure of financial openness developed by Chinn and I to (2008),which we
introduced above. Likewise, borrowing from internal sources should be easier in countries
with higher financial development. We proxy the latter using the level of domestic credit to
private sector (as share of GDP), reported by the World Development Indicators. Sovereign
wealth funds, in particular those aimed at stabilizing the fiscal stance, are increasingly popular
and we control in our regressions with the instrumented dummy variable approach discussed
above.
Institutional determinants of fiscal performance
Fiscal performance is co-determined by political conditions. Among the latter we have
include political participation, which we proxy with the Polity2 measure of democracy
compiled by the Polity IV Project 2016. Political budget cycles are recognized in many
democracies, as politicians use fiscal policy for reelection purposes. An alternative would be
to use the ICRG democracy index, which differs in terms of country and time coverage.
Democratic horizontal accountability requires that checks and balances and oversight
mechanisms are in place to exercise control over government budgets. We use as proxy
variable the measure of checks and balances developed by World Bank (2016). This index is a
quantitative measure of the institutional constraints faced by authorities in the operation of
the government. In addition to political variables, we include measures of the perceived
political stability of government, as measured by the ICRG stability index, and expect it to be
positively correlated to fiscal outcomes. Alesina and Tabelini (1990) provide several
explanations for the observation that fiscal performance is worse in unstable countries.
Exchange-rate and monetary regimes
Fiscal policy does not operate in isolation from other macroeconomic policies,
particularly monetary and exchange regimes. A standard result in the literature is that the
stabilizing effects of fiscal policy are higher the more inflexible is the exchange-rate regime.
Therefore we use two variables to address this issue: dummy variables for de jure and de facto
measures of fixed exchange-rate regimes and for membership of a monetary union.
28
Revenue Instability
Lack of revenue forecastability could be one of the reasons behind fiscal procyclicality.
It is difficult for policymakers to predict the exact timing of the business cycle and the
associated levels of revenue collection. Forecasting revenues is also difficult when the tax base
is volatile or when revenue is largely from natural resource rents that on volatile
international prices of commodities. Policymakers determine fiscal policy under a veil of
ignorance about the state of the economy. Talvi and Végh (2005) argue that finance ministers
of all countries tend to underestimate fiscal revenues to avoid political spending pressures. On
the other hand, Frankel (2011) found empirical evidence for 33 advanced and emerging
market countries that official forecasts of the budget balance and GDP growth are overly
optimistic. We proxy this lack of forecastability using a measure of fiscal revenue instability,
namely the three-year rolling coefficient of variation of tax revenue as share of GDP (see
Appendix Table 2).
External Shocks
External shocks impact fiscal performance significantly. We control for three
structural features observed in many developing economies that raise vulnerability of their
public finances to foreign shocks. Worker remittances can induce cycles in public finances via
tax revenue. We also control for the role of natural resource revenues, as well as for the
concentration of exports (as exports are more concentrated, commodity price fluctuations
become relatively more important and can induce significant cyclicality in revenues and fiscal
balances).
Cyclical Conditions
We control for temporary cyclical conditions that affect fiscal performance,
considering three potential causes of fiscal performance: domestic business cycles, resource-
rent cycles, and price instability. We use the measures discussed in the previous section.
Business cycles and price instability (inflation) can influence fiscal outcomes by affecting the
domestic tax base. Resource-rent cycles arising primarily from commodity price fluctuations
and affect several categories of revenue inn commodity-exporting countries.
Overall Development
Overall development level is measured by per capita GDP. The data is from World
Development Indicators (2017). We also include the population dependency ratio. As
discussed above, an important determinant of the dynamics of government expenditures is
the change in the demand for government spending of the young and the elderly, for childcare,
education, pensions, and health.
In the following subsections we present the econometric results. Note that we exclude
models for revenue rules given that only eight countries have implemented such rules and
their experience is not informative for emerging countries since six of them are developed
29
economies.7 Our econometric results are presented as follows: in all cases we report a “base
model” including the variables found to be important determinants of fiscal performance in
previous literature. We next report the results of a “clean model”, i.e, the base model less
those variables that in our sample of counties and years are found to be statistically and
economically irrelevant. We then provide models testing separately the role of three types of
rules: budget balance, debt and expenditure rules. We test the rules separately to avoid
collinearity. We finally extend the models to allow for differential effects in small states and in
Latin American countries.
7.3 Main Results on Procyclicality of Government Expenditures
First we study the role of fiscal rules in reducing the procyclicality of government
expenditures; a negative estimated coefficient would indicate that rules are effective. Table 4
reports the main results for our estimated panel-data, instrumental-variable estimation for
the procyclicality of government expenditures. The results for Model 1 (our base model)
strongly justify our empirical approach. First, there is significant inertia in procyclicality, thus
justifying the use of a dynamic model. Arellano-Bond type tests indicate that after including
two lags, the model is properly specified. The first two lags of the dependent variable are
significant and exhibit opposite signs indicating that the dynamics are highly persistent and
quite cyclical. Our results for the control variables replicate in a more formal environment
those of previous studies. Among the key results we find that:
Countries at higher higher income levels exhibit lower degrees of procyclicality. The
estimated coefficient of -0.335 indicates that the average small state (with a per capita
GDP that exceeds that of larger states, as reported in Table 2) would benefit from
around 40% less procyclicality than the average large economy.
Economies with more stable governments have lower procyclicality. This would
support Rogoff’s (1990) hypothesis that political business cycles can have a
deleterious effect on government finances. The parameter is nevertheless small and
the differential impact of small states vis-à-vis larger economies is negligible.
A more open capital account is associated with higher procyclicality but the
differential effect between small states and larger economies is close to zero since
there are only minor differences in financial openness.
Price instability plays a significant role in inducing higher procyclicality. Halving
annual inflation from 10% to 5% lowers procyclicality by five percentage points,
which is a sizable effect, considering that given that the sample average inflation is
9.9%.
7 Countries that have long-standing revenue rules include Australia (1996-2015), the Netherlands (1992-2015), and Kenya (1997-2015). The rest of the countries, which include Denmark, France, Lithuania, Iran, and Belgium, have had revenue rules for around a decade or less.
30
Reducing instability of government revenue by 50% – for example by adopting
multiannual budgets or implementing stabilization funds – eliminates half of the
observed procyclicality in government expenditures.
Not surprisingly, the business cycle is a key determinant of fiscal procyclicality: the
estimated parameter is large and statistically significant.
We found no effects of resource rents, worker remittances, the exchange regime, and
export concentration on procyclicality, and we therefore drop these variables from the
subsequent analysis.
Models 3 to 5 in Table 4 test for the direct impact of the three fiscal rules on
procyclicality. Estimated parameters for variables other than fiscal rules do not change
significantly (in size or statistical significance) in comparison to the Model 2.
All measures of fiscal rules used here (and in subsequent tables) are the instrumented
or fitted values estimated in stage 2, as reported in section 6 and Appendix Table 3. All
estimated parameters are negative, indicating that rules have stabilizing effects on
government expenditures, reducing procyclicality of government expenditure. However, only
the result for expenditure rules is statistically significant (with a p-value lower than 5%) and
economically large. The latter coefficient estimate suggests that countries implementing an
expenditure rule could reduce expenditure procyclicality by 40%.
Both from our analysis of different subcategories of rules (in Column (A), Table 1) and
the review of the analytical literature (in section 3), we expect that the effects of general rules
(such as our measures of balanced budget, debt, and expenditure rules) on government
expenditure cyclicality is ambiguous. Our results are in line with this expected ambiguity of
general rules: we find that two rules have no effects on expenditure cyclicality, while the third
rule has a significant and large negative impact on cyclicality.
In what follows we focus only on expenditure rules. First we interact the expenditure
fiscal rule with our population dummy (taking value 1 if the country has less than 1.5 million
inhabitants) to test for differential effects of rules in small states. Model 6 reports that the
estimated parameter is insignificant, thus indicating that there are no differential effects of an
expenditure rule for a small state. In Model 7 we test the proposition that the stabilizing
effects of an expenditure rule are different in Latin American economies, compared to other
regions. Our results reject the hypothesis as the parameter turns out to be insignificant.
31
Table 4 Determinants of the Procyclicality of Government Expenditure
------------------------------------------------------------------------------------------------------------------------ Base Model Clean Model BBR Model DR Model ER Model ER SS ER LAC ER SWF
(1) (2) (3) (4) (5) (6) (7) (8) ------------------------------------------------------------------------------------------------------------------------ 1st Lag Expenditure 0.639*** 0.665*** 0.659*** 0.664*** 0.652*** 0.637*** 0.652*** 0.677*** Procyclicality (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 2nd lag Expenditure -0.203*** -0.195*** -0.190*** -0.188*** -0.194*** -0.187*** -0.193*** -0.221***
Procyclicality (0.003) (0.002) (0.003) (0.003) (0.004) (0.005) (0.003) (0.001) Development -0.335** -0.210* -0.208* -0.205* -0.128 -0.132 -0.132 -0.0829 Level (0.028) (0.078) (0.070) (0.067) (0.273) (0.273) (0.248) (0.550) Government -0.0284 -0.0303* -0.0296 -0.0298* -0.0285 -0.0261 -0.0293 -0.0256 Stability (0.110) (0.089) (0.117) (0.096) (0.142) (0.183) (0.128) (0.220) Business Cycles 6.000** 6.800*** 6.860*** 6.910*** 6.593*** 6.358*** 6.568*** 6.074*** (0.010) (0.005) (0.005) (0.004) (0.006) (0.007) (0.007) (0.010) Financial 0.533*** 0.401** 0.406** 0.406** 0.522*** 0.493** 0.534*** 0.565*** Openness (0.002) (0.011) (0.048) (0.022) (0.005) (0.010) (0.003) (0.003) Price 0.812 0.710 0.711 0.730* 0.710 0.697 0.700 0.708 Instability (0.108) (0.102) (0.100) (0.091) (0.111) (0.121) (0.113) (0.101) Revenue 0.912** 0.826** 0.810* 0.808* 0.550 0.509 0.576 0.546 Instability (0.017) (0.046) (0.054) (0.058) (0.188) (0.222) (0.177) (0.196) Exports -0.241 -0.330** -0.325** -0.326** -0.332** -0.332** -0.334** -0.230 Concentration (0.134) (0.031) (0.038) (0.039) (0.035) (0.033) (0.036) (0.157) Workers 0.0470 Remittances (0.151) Fixed Exchange -0.120 Regime (0.225) Resource Rents 0.257 Cycles (0.298) Dependency Ratio 1.124 (0.174) Budget Bal. Rule -0.0537 (0.905) Debt Rule -0.0562 (0.944) Expenditure Rule -1.434** -1.472** -1.362* -1.970** (0.034) (0.031) (0.054) (0.034) Expenditure Rule* 1.469 Small State (0.519) Expenditure Rule* -0.496 LAC (0.694) Sovereign Wealth -0.944* Fund (0.068) Expenditure Rule* 3.634 Sov. Wealth Fund (0.187) Constant -0.115 4.228*** 4.173*** 4.156*** 3.614** 3.649** 3.668** 2.499 (0.973) (0.005) (0.007) (0.006) (0.026) (0.023) (0.025) (0.195) ------------------------------------------------------------------------------------------------------------------------ Observations 408 414 414 414 414 414 414 414 Number of Countries 111 111 111 111 111 111 111 111 Arellano Bond Tests -4.10*** -4.54*** -4.50*** -4.58*** -4.47*** -4.42*** -4.45*** -4.73*** -1.36 -1.59 -1.61 -1.65* -1.37 -1.33 -1.30 -1.27 Sargan Test 20.68 21.55 22.14 22.21 21.84 21.39 21.81 21.07 ------------------------------------------------------------------------------------------------------------------------ Note: Dependent variable is the 10-year rolling correlation of HP-filtered expenditures of general government to HP-filtered GDP. The econometric method used is System Dynamic Panel-data Models with Instrumental Variables. Fiscal rules and Sovereign Wealth Fund are instrumented using the results in section 6. p-values in parentheses. * p<0.1, ** p<0.05, *** p<0.01.
32
Finally, in Model 8 we investigate whether our expenditure rule instrument could be
representing another fiscal institution, namely a sovereign wealth fund. The coefficients of
both the expenditure rule and the sovereign wealth fund are large, negative and statistically
significant, indicating that both fiscal institutions are important in reducing fiscal
procyclicality. Furthermore, the estimated coefficient for the expenditure rule in Model 5 is
smaller than that in Model 8 and the interaction term is insignificant, which leads to conclude
that our initial result (Model 5) on the stabilizing effect of an expenditure rule is robust.
7.4 Main Results on Procyclicality of Fiscal Balance
Now we replicate the previous analysis to assess the procyclicality of the fiscal
balance. Note that here the stabilizing role of fiscal rules would be reflected in an increase in
the cyclicality measure, i.e., the government budget balance ought to increase whenever the
economy expands to be considered a countercyclical policy. The results for Base Model 1 in
Table 5 indicate that there is significant inertia in fiscal balances, as expected. The opposite
signs of the first two lags of the dependent variable reveal the complex dynamics of an
adjustment that is both persistent and cyclical. Again, Arellano-Bond type tests indicate that
two lags must be included to have a properly specified model.
With regards to control variables, the results are the following:
Procyclicality of fiscal balances does not depend on development levels (the estimated
parameter is negative but statistically insignificant). This result invalidates the notion
that poor, small states are in worse condition to manage their fiscal stance.
Countries with high levels of checks and balances on the operation of the government
tend to have a more stable fiscal stance on average (the estimated parameter is
positive and statistically significant). Again this would be in line with Rogoff’s (1990)
political business cycle model, where incumbents cut taxes or increase spending prior
to elections to increase their chance of reelection but are forced to subsequently
increase taxes or cut spending once reelected, thus inducing fiscal procyclicality.
In line with the literature we find a positive association between business cycles and
price instability and the procyclicality of fiscal balances, but the estimated coefficients
are statistically insignificant at 90% confidence. This could be the results of averaging
over five-year periods which dampens the variance of regressors. The positive
parameter estimated for the effect of having a fixed exchange rate on fiscal balances
would be also in line with this explanation because a credible fixed exchange would
lower the depreciation risk and thereby the cost of funding. Higher dependency ratios
are associated with higher procyclicality.
33
Table 5 Determinants of the Procyclicality of Fiscal Balance
------------------------------------------------------------------------------------------------------------------------ Base Model Clean Model BBR Model DR Model ER Model BBR LAC DR LAC ER LAC
(1) (2) (3) (4) (5) (6) (7) (8) ------------------------------------------------------------------------------------------------------------------------ 1st Lag Budg. Bal. 0.522*** 0.487*** 0.471*** 0.458*** 0.489*** 0.477*** 0.465*** 0.477*** Procyclicality (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 2nd lag Budg. Bal -0.164** -0.176*** -0.164*** -0.173*** -0.183*** -0.174*** -0.174*** -0.169*** Procyclicality (0.015) (0.005) (0.009) (0.005) (0.002) (0.004) (0.004) (0.004) Development -0.0344 -0.0252 0.00853 -0.00418 -0.0381 -0.000725 -0.00716 0.0121 Level (0.823) (0.879) (0.956) (0.981) (0.806) (0.996) (0.966) (0.938) Checks and 0.0229** 0.0211** 0.0225** 0.0218** 0.0199* 0.0216** 0.0219** 0.0226* Balances (0.022) (0.030) (0.016) (0.024) (0.089) (0.022) (0.022) (0.062) Financial 0.0955 0.0792 0.0756 0.0732 0.0892 0.0856 0.0869 0.0959* Development (0.164) (0.170) (0.175) (0.201) (0.157) (0.126) (0.134) (0.085) Fixed Exchange 0.250** 0.241** 0.244** 0.243** 0.240** 0.249** 0.248** 0.254** Regime (0.030) (0.032) (0.032) (0.033) (0.029) (0.028) (0.026) (0.023) Revenue -0.750** -0.826** -0.794** -0.813** -0.790** -0.879** -0.852** -0.896** Instability (0.012) (0.027) (0.021) (0.021) (0.028) (0.020) (0.021) (0.015) Dependency Ratio 2.448*** 1.875*** 1.836*** 1.738** 1.864*** 1.500** 1.503** 1.241** (0.000) (0.002) (0.007) (0.015) (0.001) (0.018) (0.031) (0.043) Sovereign Wealth 0.303 0.544* 0.454 0.481 0.454 0.546 0.549* 0.556 Fund (0.468) (0.079) (0.180) (0.121) (0.176) (0.103) (0.098) (0.103) Business Cycles 3.218 (0.157) Price 0.337 Instability (0.341) Financial 0.0580 Openness (0.791) Workers -0.0361 Remittances (0.214) Resource -0.0157 Rents (0.737) Exports 0.132 Concentration (0.407) Budget Bal. Rule 0.239 -0.107 (0.470) (0.789) Budget Bal. Rule* -4.510* Small State (0.072) Debt Rule 0.439 -0.0497 (0.428) (0.936) Debt Rule* -2.147 Small State (0.542) Expenditure Rule 0.374 -0.321 (0.526) (0.614) Expenditure Rule* 5.586 Small State (0.209) Budget Bal. Rule* 0.919 LAC (0.130) Debt Rule*LAC 1.313 (0.144) Expenditure Rule* 2.864** LAC (0.025) Constant -10.75*** -7.735*** -7.856*** -7.339*** -7.653*** -6.436*** -6.400*** -5.516** (0.000) (0.000) (0.000) (0.001) (0.000) (0.003) (0.006) (0.011) ------------------------------------------------------------------------------------------------------------------------ Observations 389 415 415 415 415 415 415 415 Number of Countries 108 110 110 110 110 110 110 110 Arellano Bond Tests -3.71*** -4.09*** -3.95*** -3.93*** -4.18*** -4.13*** -4.08*** -4.12*** 0.20 0.35 0.28 0.27 0.46 0.25 0.29 0.19 Sargan Test 19.39 25.10* 23.86 23.68 23.76 23.08 23.80 22.25 ------------------------------------------------------------------------------------------------------------------------
Note: Dependent variable is the 10-year rolling correlation of HP-filtered fiscal balance to HP-filtered GDP. The econometric method used is System Dynamic Panel-data Models with Instrumental Variables. Fiscal rules and Sovereign Wealth Fund are instrumented using the results in section 6. p-values in parentheses. * p<0.1, ** p<0.05, *** p<0.01.
34
Two additional results are in line with those reported for expenditure procyclicality.
First, we confirm that sovereign wealth funds not only have a stabilizing effect on
expenditures but also on fiscal balances.8 Second, we found that higher instability in
fiscal revenues not only induces procyclicality of expenditures but also the
procyclicality of fiscal balances.
Turning now to the role of rules in the procyclicality of fiscal balances, in Model 3 the
evidence indicates that budget balance rules – which ought to be directly linked to the
stability of the government stance – in fact have a positive but statistically insignificant
estimated coefficient. Likewise, the estimated coefficients for the debt and expenditures rules
in Models 4 and 5 are also positive but statistically insignificant. Note also that the estimated
coefficients for the rest of variables do not change in any manner, suggesting that this result
does not arise from collinearity problems. Our models also test the hypothesis that fiscal rules
in small states would have a differential effect on the procyclicality of the fiscal stance. It can
be seen that all of the parameters are statistically zero, except for budget balance rules which
display a negative estimate, implying that in small economies this type of rule could backfire.
This reinforces the above conclusion that general fiscal rules do not contribute to more stable
fiscal balances and that this effect in general is not conditioned by economic size.
We also find evidence that expenditure rules can have a significant positive effect on
the stability of fiscal balances on the 11 Latin American countries that, at one point or another
of their recent history, have had rules in operation. As shown in Models 6, 7, and 8 of Table 5,
the interactions between our instrumented fiscal rules and a regional dummy for Latin
America (LAC) are all positive but only the coefficient of the expenditure is highly significant.
Leaving aside the issue of statistical significance, note that the point estimates of these effects
are also asymmetrical being expenditure rules more effective in reducing the procyclicality of
fiscal balances than budget balance rules. The evidence on the beneficial effect of rules in
Latin America is entirely derived from large economies (the smallest is Panama with 3.5
million inhabitants) and thus cannot be extended directly to small states. Nevertheless, to the
extent that Latin American economies share similar institutional setups and development
levels, the results is indicative of the potential gains of fiscal rules in this region.
7.5 Main Results on Fiscal Balances
We now turn to the results for our models on the role of fiscal rules in affecting fiscal
balances and sustainability. We deem an economy to be more sustainable the lower is the
fiscal deficit and, therefore, a negative estimated coefficient would indicate that rules are
effective. As shown in Table 6, and in agreement with the previous models, there is a high
level of inertia in fiscal balances. Understandably, adjusting fiscal imbalances –particularly,
high deficits—requires a lot of effort on the part of authorities as well as political muscle, and
it usually takes time. The main results for our standard controls are as follows:
8 There are only nine small economies with a SWF (Bahrain, Brunei, Gabon, Kuwait, Oman, Qatar, Trinidad & Tobago, United Arab Emirates). Interaction dummies are statistically insignificant.
35
We found the standard result in the literature that more stable countries in political
terms as well as those with fixed exchange rates tend to have more conservative fiscal
policies and, on average, higher fiscal balances. However, we found no effects of the
development level itself, indicating that a balanced budget is not a privilege of
advanced economies.
Business and resource-rent cycles play a significant role in affecting fiscal balances,
even if in a transitory manner. We found no significant effects of fiscal revenue
instability on fiscal balances: the latter result is an interesting finding.
The empirical evidence indicates that countries with fixed exchange regimes tend to
have lower fiscal deficits or higher surpluses, a result that could be in part explained
because resource-rich economies (e.g., oil exporters) usually have large fiscal
surpluses and have historically preferred to peg their currencies either to the British
pound or the US dollar.
We find that larger inflows of worker’s remittances deteriorate fiscal balances. This is
in line with the analytical results of Chami et al. (2009) which found that, by
increasing consumption and the revenue base, remittances may allow the government
to carry more debt or incur in more expenditures.
Models 3, 4, and 5 in Table 6 test whether fiscal rules have a positive impact on fiscal
balances: the estimated coefficients for all types of rule are positive and statistically
significant. These econometric results confirm the predictions of our theory, as summarized in
Table 1. According to our theoretical model, all fiscal rules should lower the fiscal deficit or,
conversely, improve the fiscal balance. This is precisely our empirical result. It is, therefore,
directly to conclude that having a fiscal rule in place leads to improve fiscal balances and
thereby supports more sustainable fiscal policies. These models also test for possible effects
of fiscal rules according to the size of the economy; we found negative but statistically
insignificant effects, except for the debt rule.
Models 6, 7, and 8 indicates that there are insignificant differential effects of fiscal
rules in Latin American countries vis-à-vis the rest of the world. For non-Latin American
countries we find that all rules have the expected positive results of improving the fiscal
balance but the opposite result and with similar magnitude – is found for Latin American
countries. The latter are very imprecisely estimated, as shown in relatively low p-values, but
one has to bear in mind that there is a small number of countries in the region with fiscal
rules. One possible explanation is that LA countries had initially imposed budget rules with
strict fiscal surpluses and later had relaxed the requirement, aiming at budgets in balance or
in deficits. This is Chile’s experience, for example. In such case, a negative correlation could be
observed between the budget balance rule and fiscal balances. Alternatively, the rule had not
been binding and credible and had not been obeyed (non-compliance). The available data on
36
de jure fiscal rules preclude us from investigate the issue. However, we return to this issue
below using data on compliance.
Table 6 Determinants of Fiscal Balance
------------------------------------------------------------------------------------------------------------------------ Base Model Clean Model BBR Model DR Model ER Model BBR LAC DR LAC ER LAC (1) (2) (3) (4) (5) (6) (7) (8) ------------------------------------------------------------------------------------------------------------------------ 1st Lag Fiscal 0.376*** 0.508*** 0.526*** 0.515*** 0.513*** 0.542*** 0.510*** 0.503*** Balance (0.006) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 2nd lag Fiscal -0.109* -0.143** -0.131** -0.139*** -0.132** -0.128** -0.138** -0.132** Balance (0.062) (0.013) (0.016) (0.009) (0.019) (0.022) (0.015) (0.027) Development 0.374 -0.0334 -0.524 -0.396 -0.504 -0.677 -0.439 -0.532 Level (0.719) (0.961) (0.419) (0.537) (0.478) (0.341) (0.518) (0.468) Government 0.277** 0.363*** 0.370*** 0.360*** 0.356*** 0.401*** 0.341*** 0.359*** Stability (0.026) (0.003) (0.003) (0.003) (0.004) (0.001) (0.004) (0.004) Fixed Exchange 1.874** 2.591*** 3.088*** 2.723*** 2.611*** 3.052*** 2.559*** 2.605*** Regime (0.018) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Business Cycles 48.65*** 67.70*** 69.09*** 67.15*** 68.05*** 69.08*** 68.28*** 67.91*** (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Cycles in 4.772*** 4.531*** 5.214*** 4.831*** 4.971*** 5.020*** 4.601*** 4.674*** Resource Rents (0.002) (0.005) (0.002) (0.003) (0.002) (0.002) (0.003) (0.003) Price 6.523* 7.588** 8.859** 8.949** 8.490** 9.058** 8.094** 8.246** Instability (0.064) (0.039) (0.021) (0.013) (0.026) (0.033) (0.037) (0.049) Workers -0.316 -0.477** -0.410** -0.503** -0.421* -0.396* -0.479** -0.443* Remittances (0.104) (0.036) (0.048) (0.019) (0.051) (0.098) (0.041) (0.063) Dependency Ratio 0.364 (0.961) Exports 0.249 Concentration (0.788) Revenue -3.020 Instability (0.269) Financial 0.490 Openness (0.709) Budget Bal. Rule 5.595** 7.387** (0.031) (0.014) Budget Bal. Rule* -21.05 Small State (0.103) Debt Rule 7.985** 9.890** (0.022) (0.015) Debt Rule* -39.83** Small State (0.044) Expenditure Rule 6.774* 6.617 (0.095) (0.136) Expenditure Rule* -20.64 Small State (0.524) Budget Bal. Rule* -8.394 LAC (0.139) Debt Rule*LAC -11.83 (0.126) Expenditure Rule* -2.272 LAC (0.832) Constant -11.73 -5.644 -2.446 -3.237 -2.251 -1.368 -2.681 -2.053 (0.689) (0.347) (0.661) (0.567) (0.705) (0.823) (0.657) (0.744) ------------------------------------------------------------------------------------------------------------------------ Observations 406 477 477 477 477 477 477 477 Number of Countries 113 121 121 121 121 121 121 121 Arellano Bond Tests -1.15 -2.88*** -2.90*** -2.95*** -2.90*** -2.94*** -2.89*** -2.79*** 0.07 -0.56 -0.60 -0.52 -0.60 -0.68 -0.48 -0.61 Sargan Test 63.92 49.85 50.62 47.70 57.10 49.59 44.68 48.28 ------------------------------------------------------------------------------------------------------------------------ Note: Dependent variable is fiscal balance (as share of GDP). The econometric method used is System Dynamic Panel-data Models with Instrumental Variables. p-values in parentheses. Fiscal rules are instrumented using the results in section 6. * p<0.1, ** p<0.05, *** p<0.01
37
7.6 Main Results on Government Debt
Our last set of results assesses the effects of fiscal rules on the stock of government
debt, which we collect in Table 7. Clearly, a negative estimated coefficient would indicate that
rules are effective in reducing debt levels and, thereby, in improving fiscal sustainability. It
can be seen that, contrary to the rest of fiscal outcome indicators, we have included only one
lag of the dependent variables as the dynamics in this case are simpler: there is significant
persistence, as is expected from a slow adjusting stock variable, but no evidence of cycles
whatsoever. Given that variable of interest is expressed as share of real GDP, debt levels
indirectly indicate debt sustainability: since tax revenue is positively correlated with GDP
levels, a lower level of debt to GDP ought to be easier to service and, therefore, more
sustainable.
With regards to the standard set of controls used in this paper, the following elements
stand out:
Development levels do not appear to correlate in any form to government debt.
However, financial development levels are positively associated with debt levels,
perhaps reflecting that when domestic financial markets are deeper, the government
finds it easier or less expensive to recourse to debt.
We found that countries endowed with more natural resources do not seem to carry
more debt. Although there are examples of countries which have mismanaged natural
resource revenues (mainly during the oil boom of the 1970s), our evidence is
inconsistent with the notion of “boom based borrowing” and “debt overhang” as
discussed by Manzano and Rigobon (2007).
Higher government stability is associated with lower government indebtedness, as
indicated by the estimated negative parameter, which is only marginally insignificant
but in subsequent regression shows up invariably significant.
A higher dependency ratio is very negatively associated to debt levels.
As expected, higher inflation levels tend to be associated with lower debt levels for
two reasons. First, governments can substitute inflationary tax for debt. Second,
governments find it costlier to finance debt in high inflation environments. We also
obtain the customary results that debt levels are lower in countries with fixed
exchange regimes.
We now examine the results for fiscal rules in Table 7. Models 3, 4, and 5 test for the
direct effect of fiscal rules: all estimated parameters are insignificantly different from zero.
This would indicate that having a debt rule in place is not tantamount to debt control or
improved sustainability, an issue to which we return below. In Model 3 we obtain a negative
and significant differential effect of a budget balance rule on debt levels in small states,
suggesting that small economies benefit from such rule. We found no differential effects of
38
expenditure and debt rules in Latin America countries vis-à-vis the rest of the world, but
found that budget balance rules tend to be associated with higher debt levels in Latin
American economies.
Table 7 Determinants of Government Debt
------------------------------------------------------------------------------------------------------------------------ Base Model Clean Model BBR Model DR Model ER Model BBR LAC DR LAC ER LAC (1) (2) (3) (4) (5) (6) (7) (8) ------------------------------------------------------------------------------------------------------------------------ 1st Lag Government 0.688*** 0.709*** 0.720*** 0.684*** 0.714*** 0.709*** 0.696*** 0.717*** Debt (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Development 8.027 7.185 5.105 9.623 8.738 8.309 9.659 8.766 Level (0.217) (0.245) (0.411) (0.197) (0.207) (0.152) (0.151) (0.222) Government -0.759 -0.771 -0.871 -0.718 -1.055 -0.987 -0.491 -0.991 Stability (0.604) (0.522) (0.466) (0.579) (0.464) (0.398) (0.704) (0.480) Dependency Ratio -167.7* -189.1** -161.0** -202.3** -189.3*** -179.5** -203.0*** -188.7*** (0.053) (0.012) (0.032) (0.015) (0.007) (0.012) (0.008) (0.007) Financial 18.38** 20.77*** 20.22*** 20.61*** 19.57*** 20.52*** 20.88*** 19.63*** Development (0.020) (0.000) (0.000) (0.002) (0.002) (0.000) (0.001) (0.002) Price -271.3** -263.8*** -251.2** -281.2*** -267.8** -261.4*** -258.5** -263.7*** Instability (0.013) (0.007) (0.015) (0.008) (0.010) (0.006) (0.012) (0.009) Fixed Exchange -19.67** -19.77*** -19.16** -20.11*** -19.38*** -20.86*** -19.40*** -19.68*** Regime (0.026) (0.007) (0.016) (0.009) (0.009) (0.007) (0.010) (0.009) Financial -24.08* -26.05** -16.86 -31.15*** -22.20 -26.17** -31.39*** -23.31 Openness (0.070) (0.029) (0.186) (0.007) (0.139) (0.022) (0.007) (0.101) Resource -0.975 Rents (0.744) Workers -1.005 Remittances (0.658) Exports 8.844 Concentration (0.296) Budget Bal. Rule -11.85 -35.21 (0.697) (0.270) Budget Bal. Rule* -256.8*** Small State (0.003) Debt Rule 19.35 3.420 (0.658) (0.945) Debt Rule* 114.0 Small State (0.616) Expenditure Rule -25.49 -28.18 (0.476) (0.430) Expenditure Rule* -117.0 Small State (0.630) Budget Bal. Rule* 65.76* LAC (0.099) Debt Rule*LAC 80.26 (0.189) Expenditure Rule* 22.73 LAC (0.847) Constant 654.1* 794.7*** 691.3** 832.2*** 783.9*** 750.0*** 831.4*** 781.0*** (0.051) (0.004) (0.012) (0.005) (0.002) (0.004) (0.003) (0.002) ------------------------------------------------------------------------------------------------------------------------ Observations 366 390 390 390 390 390 390 390 Number of Countries 116 116 116 116 116 116 116 116 Arellano Bond Tests -1.23 -1.14 -1.22 -1.12 -1.14 -1.17 -1.11 -1.16 Sargan Test 27.51 32.09 31.58 34.00 30.91 30.26 33.99 31.03 ------------------------------------------------------------------------------------------------------------------------ Note: Dependent variable is Government Gross Debt (as share of GDP). The econometric method used is System Dynamic Panel-data Models with Instrumental Variables. p-values in parentheses. Fiscal rules are instrumented using the results in section 6. * p<0.1, ** p<0.05, *** p<0.01
39
Three considerations on the role of general de jure fiscal rules are important to
highlight at this point. First, having any general fiscal rule in place does not ensure that it will
contribute to a particular fiscal policy objective. In the absence of data on particular sub-types
of rules, the contribution of general rules to macroeconomic stabilization is uncertain. Second,
having a fiscal rule in place does not necessarily imply that such rule is optimally designed or
even appropriate for a country in a particular situation. For example, fiscal rules imposing
non-binding restrictions or with non-credible design (e.g., without properly designed escape
clauses) would have no positive effects (and perhaps even negative effects) on fiscal
outcomes. Unfortunately, our dataset does not allow us to address this important issue and
our econometric models implicitly assume that de jure rules are well designed. Third, having a
rule in place does not necessarily mean that it is properly enforced by the authorities.
Countries may choose to bypass or ignore altogether de jure fiscal rules, implying partial or
full non-compliance. This leads to exploring the empirical relevance of compliance of
countries in their implementation of de jure rules, which we do in the following section.
8 De Facto and de Jure Rules
In this section we first address the issue of de jure vs. de facto fiscal rules. Two
practical considerations arise when using measures of compliance. First, we use the slightly
more restricted data on fiscal rule compliance collected by Belinga et al. (2017) in our
empirical models; out of the more than 80 countries with de jure national fiscal rules in 2015
only around 70 countries have data on compliance or de facto rules. Second, the data is likely
to be noisy since it corresponds to the assessment made by country economists and other
experts on the degree of compliance of the rules in any given country and for every year in the
sample. Noise inevitably arises because the data corresponds to the judgment of different
individuals with varying degrees of information and, in most cases, it would be memory-based
or the product of ex-post analysis rather than direct witnessing of events. Data on compliance
has been coded as 1 if there was compliance on any given year and zero otherwise. The latter
would include countries without fiscal rules as well as those that do not comply with the rules
they chose to implement.
We replicate the models in Tables 4 to 7 using the data on compliance instead of de
jure rules. We include in our models both the rules by themselves as well as in interaction
with the dummy variable for Latin America, the latter interacted only for budget balance and
expenditure rule compliance since there is no data for debt compliance in Latin American
economies. We also include the dummy for small states. The results are presented in Table 8,
where we omit the results for standard controls for space reasons and focus only on the de
facto rules. Column 1 presents the result on expenditure procyclicality. It can be seen that all
de facto rules and the small-state dummies are statistically insignificant. The result is very
similar to that of de jure rules, except that in the latter the expenditure rule was significantly
counter-cyclical. With regards to the procyclicality of fiscal balance, we found that de facto
rules have no direct impact, as was the case of de jure rules: estimated parameters are
statistically insignificant (see Column 2). Likewise, the interaction between de facto rules and
40
small-state dummies are also statistically insignificant, suggesting that small-size economies
are not intrinsically different than their larger counterparts vis-à-vis procyclicality.
Table 8 Estimated Results of Models using Fiscal Rule Compliance
------------------------------------------------------------------------------------------------------------------------ Exp. Pro. Bal. Pro. F. Balance Debt Exp. Pro. Bal. Pro. F. Balance Debt (1) (2) (3) (4) (5) (6) (7) (8) ------------------------------------------------------------------------------------------------------------------------
Standard Controls Omitted Compliance with -0.0340 0.192 3.099*** -9.704* -0.0569 0.223 2.971*** -9.917 BR (0.766) (0.113) (0.000) (0.097) (0.689) (0.102) (0.000) (0.101) Compliance with 0.245 -0.210 -1.595* -7.610 0.256 -0.258* -1.462* -4.269 DR (0.388) (0.200) (0.074) (0.343) (0.367) (0.095) (0.081) (0.622) Compliance with 0.00854 0.173 1.110 12.33 -0.0732 0.266* 0.208 4.771 ER (0.958) (0.265) (0.232) (0.284) (0.676) (0.061) (0.836) (0.562) Compliance with -0.200 -0.0356 3.011 -41.76 BR*Small State (0.592) (0.915) (0.358) (0.372) Compliance with 1.302 -161.4 DR*Small State (0.905) (0.413) BR Compliance* 0.131 -0.0969 1.032 5.799 LAC (0.568) (0.616) (0.443) (0.483) ER Compliance* 0.550 -0.566 3.466 100.1*** LAC (0.189) (0.382) (0.155) (0.000) DR Compliance* -2.117 -5.442 LAC (0.166) (0.711) Constant 2.749 -7.539*** -4.220 562.4 2.594 -8.445*** -2.341 625.3 (0.160) (0.001) (0.500) (0.199) (0.201) (0.000) (0.725) (0.154) ------------------------------------------------------------------------------------------------------------------------ Observations 414 415 477 279 414 415 477 279 Number of Countries 111 110 121 115 111 110 121 115 ------------------------------------------------------------------------------------------------------------------------ Note: The econometric method used is System Dynamic Panel-data Models with Instrumental Variables. Fiscal rules and
Sovereign Wealth Fund are instrumented using the results in section 6.p-values in parentheses, * p<0.1, ** p<0.05, *** p<0.01
Column 3 of Table 8 collects the results for the models testing the effects of de facto
rules on fiscal balances. It can be seen that compliance with the budget balance rule actually
improves significantly the fiscal stance, as expected. This effect was not visible when using the
de jure rule, and it is a significant contribution of the new dataset. The de facto expenditure
rule has no discernible effect on the fiscal stance, but the debt rule has a unexpected negative
correlation to the fiscal balance. Note that because we cannot instrument compliance, we can
at best analyze our econometric results in terms of correlations and not of causality, as
discussed in section 5. The last set of results on compliance refer to the effect of rules on
government debt and are presented in column 4 of Table 8. Again, the debt rule even if
complied has no effects in lowering debt level and, on the contrary, it is the compliance with a
budget balance rule lowers the debt by a sizable amount.
We extend our analysis to test for differential effects of the facto rules on fiscal
outcomes in Latin American economies and present the results in the last four columns of
Table 8. It can be seen that de facto rules and the LAC dummy are statistically insignificant in
both procyclicality models (expenditures and fiscal balances) thus leading to the conclusion
that LAC economies are not different than other countries of the world in this regard.
Likewise, our results for fiscal balances do not change with the inclusion of the LAC
41
interaction, leading to the conclusion that there are no regional asymmetries. Finally, we
disregard the estimated model for the government debt on the grounds that there is not
enough density for ER in LAC, even if the parameters appears to be highly significant.
9 Conclusions
This paper presents world evidence on the contribution of fiscal rules to fiscal
performance. The paper starts by identifying the links between fiscal institutions, rules, and
performance measures and reviewing the relevant analytical and empirical literature on fiscal
performance, and on the potential contribution of different fiscal rules to performance. Our
analysis indicates that the impact of rules on fiscal outcomes (balance, debt, and
procyclicality) have three defining characteristics: (a) they tend to display high persistence,
thereby requiring the use of dynamic models, (b) countries are highly idiosyncratic vis-à-vis
unobservable features in their tax and revenue structures, prompting the existence of
individual country effects, and (c) there may be reverse causation and endogeneity, since
governments may decide to implement rules as a result of their fiscal stance.
Our methodology assumes that unveiling the impact of rules on fiscal performance can
be casted as the study of whether a treatment (implementing a fiscal rule) has had any
discernible effect – the treatment effect – on fiscal performance. Identifying causal effects
from observational data is often quite difficult as it depends crucially on the validity of the
implicit or explicit identification assumptions. In our context of panel data, this would indicate
the need of properly specifying the determinants of fiscal performance as well as controlling
for the potential endogeneity of the treatments. Furthermore, modelling the treatment is not
as straightforward as it might seem as it requires separating the determinants of fiscal
outcomes from those that led to the enactment of such rules. We use standard dynamic panel
models to address the first two characteristics. In order to avoid our results to be hampered
by endogeneity running from fiscal outcomes to treatments, we complement dynamic panel
data techniques with a novel approach to building a consistent instrument for the choice of
implementing a fiscal rule.
Our conclusions on the role of fiscal rules on performance are based on two
considerations. First, the inherent differences between small states and larger economies in
terms of economic structure and size, and the external shocks they face. Second, the
econometric evidence obtained for a group of 115 economies in the period 1980-2015
regarding the impact of rules on four dimensions of fiscal performance: government
expenditures procyclicality, fiscal balance procyclicality, fiscal balances and government debt.
Using the World Bank definition of a small state (population below 1.5 million) we
found that they have higher development levels but enjoy more government stability, even if
political checks and balances are slightly less developed. We also confirm that small states are
significantly more open to foreign trade, their export base is less diversified, and that natural
resources are less important. Small states are equally open in terms of financial markets but
tend to prefer fixed exchange rate systems despite receiving much larger inflows of worker
42
remittances. Despite the abovementioned differences, large and small states share similar
business cycles, inflation levels, and even similar levels of tax revenue instability. In summary,
small states share only a few characteristics with larger economies so that one should expect a
potentially different impact of fiscal rules according to economic size.
One general conclusion from the econometric evidence is that there is significant
inertia in fiscal balances, with complex dynamics and adjustments paths that are both
persistent and cyclical. Therefore, the impact of policy shocks as well as of implementing fiscal
rules or any other fiscal institution takes a relatively long time to materialize, be it in the form
of reduced procyclicality or higher sustainability. Should small states in Latin America decide
to implement fiscal rules – and should these rules be properly designed – the beneficial
results would accrue only in the medium to long run.
A second general conclusion from our evidence is that, while fiscal rules are important
instruments to improve the fiscal stance of an economy, other variables can be equal or more
important determinants of fiscal performance. Elements that improve the fiscal performance
of a country include having a properly designed and functioning system of checks and
balances in the government; achieving higher levels of political stability; fostering the
integration of the economy with globalized financial markets; and maintaining lower and
stable levels of inflation. The flip-side of these results indicate that absent these elements, the
benefits of implementing fiscal rules may be severely hampered.
A third general conclusion is that fiscal institutions other than rules also play an
important role in reducing procyclicality and/or increasing sustainability. We found that
reducing the instability of government revenues would eliminate a significant portion of fiscal
procyclicality in government expenditures and of fiscal balances. Among the alternatives
policies to stabilize revenues are, for example, adopting multiannual budgets, diversifying the
tax base away from natural resource exports or tourism, and implementing revenue-
stabilization funds. By themselves, these measures would certainly improve matters in a cost-
effective manner; implemented in conjunction with appropriate fiscal rules their impact
would be much enhanced.
A fourth general conclusion is that cyclical phenomena are quite important
determinants of the fiscal stance and, most likely, they should be dealt with outside the realm
of fiscal rules. We found that business cycles, resource cycles induced by commodity price
fluctuations, and changes in workers’ remittances have a deleterious impact on fiscal
performance. Dealing with these elements would call for instruments better equipped to deal
with transitory shocks than fiscal rules which tend to focus on medium-to-long term goals
(e.g., structural balance or debt targets) and use rather inflexible instruments such as annual
budgets. In this regard, we found that sovereign wealth funds have an important stabilizing
effect on fiscal balances and are instrumental in reducing procyclicality.
Turning now to the evidence regarding the role of fiscal rules we found that there are
some important differences in terms of effectiveness between de jure and de facto rules. The
former refer to a rule that imposes a long-lasting constraint on fiscal policy through numerical
limits on budgetary aggregates. Such constraints can be easily identified and measured;
43
furthermore, an instrumental variables procedure is used in our study to avoid biasing or
estimation as a consequence of endogeneity. Of course, compliance with such constraints is an
entirely different matter: the de facto measure of fiscal rules collected by Belinga et al. (2017)
and used in our study provides important information regarding the actual operation of the
rules but its measurement based on experts’ opinions is rather subjective.
With regards to de jure rules, a descriptive summary of results is found in the
following table. It can be seen that two rules deliver their intended primary result:
expenditure rules reduce expenditure procyclicality and budget balance rules improve fiscal
balances. Debt rules, on the contrary, do not affect debt levels in a significant manner. One
explanation for this result is that debt rules are only binding from above when indebtedness
reaches a certain target, prior to that level the rule is ineffective. Also no rule has a direct
impact on the procyclicality of fiscal balances, perhaps because rules target either one
component (expenditures) or the level of fiscal deficit and in achieving any of such targets
might induce volatile instead of controlling it. We also find that debt and budget balance rules
do not have effects outside their primary target, but expenditure rules not only reduce fiscal
procyclicality but also improve on fiscal balances.
To a large extent, the effectiveness of fiscal rules in achieving their intended objective
is independent of economic size except for a few cases: small states seem to benefit more in
terms of improved sustainability from budget balance and debt rules. We also found evidence
that, while an expenditure rule might not reduce expenditures procyclicality differently in
Latin America than in the rest of the world, it can act as a stabilizing tool for fiscal balances.
When studying the same issues but using de facto instead of de jure fiscal rules the
econometric results do not change significantly and offer only one additional insights, namely
that compliance with expenditure rules might allow Latin American countries higher levels of
government debt.
Table 10 Summary of Results
Fiscal Outcomes Budget Balance Rule Debt Rule Expenditures Rule
de jure de facto de jure de facto de jure de facto
Procyclicality of government expenditures
Is there any effect on procyclicality? No No No No Reduced Reduced Small states are more/less procyclical? No No No No No No LAC countries are more/less procyclical? No No No No No No Procyclicality of fiscal balances
Is there any effect on procyclicality? No No No No No No Small states are more/less procyclical? More No No No No No LAC countries are more/less procyclical? No No No No Less No Fiscal Balance
Do fiscal balances improve? Yes Yes Yes Yes Yes No Small states have higher/lower balances? No No Lower No No No LAC countries have higher/lower balances? No No No No No No Government debt
Is debt reduced? No No No No No No Small states have higher/lower debt? Lower No No No No No LAC countries have higher/lower? No No No No No Higher
44
References
Abdih, Y., Chami, R., Mati, A., and Gapen, M. (2009). Fiscal Sustainability in Remittance-
Dependent Economies. IMF Working Paper WP/09/190.
Adams, R., Almeida, H., and Ferreira, D. (2009). Understanding the relationship between
founder–CEOs and firm perform. Journal of Empirical Finance 16, 136–150.
Aizenman, J., and Glick, R. (2009). Sovereign Wealth Funds: Stylized Facts about their
Determinants and Governance. International Finance, 12(3), 351-386.
Aizenman, J., Gavin, M., and Hausmann, R. (2000). Optimal tax and debt policy with
endogenously imperfect creditworthiness. Journal of International Trade & Economic
Development, 9(4), 367-395.
Alesina, F., and Tabellini, G. (1990). A Positive Theory of Fiscal Deficits and Government Debt.
Review of Economic Studies, 57(3), 403-414.
Anderson, T. W., and Hsiao, C. (1981). Estimation of dynamic models with error components.
Journal of the American Statistical Association, 76(1981), 589–606.
Angrist, J. D., and Kuersteiner, G. M. (2011). Causal Effects of Monetary Shocks:
Semiparametric Conditional Independence Tests with a Multinomial Propensity Score.
Review of Economics and Statistics, 93(3), 725-747.
Angrist, J. D., Jordà, Ò., and Kuersteiner, G. M. (2013). Semiparametric Estimates of Monetary
Policy Effects Before and Since the Great Recession: String Theory Revisited. Working
Paper 2013-24, Federal Reserve Bank of San Francisco.
Angrist, J., and Krueger, A. B. (2001). Instrumental variables and the search for identification:
from supply and demand to natural experiments. Journal of Economic Perspectives 15,
69–87.
Arellano, M., and Bond, S. (1991). Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. Review of Economic Studies,
58(2), 277–97.
Barnett, M. S., and Ossowski, M. R. (2002). Operational aspects of fiscal policy in oil-producing
countries. IMF Working Paper WP/02/177.
Beck, R., and Fidora, M. (2008). The Impact of Sovereign Wealth Funds on Global Financial
Market. ECB Occasional Paper Series No 91/July.
Belinga, V., Gąska, J., Hołda, P., and Skrok, E. (2017). The Impact of Fiscal Rules on Fiscal
Performance. The World Bank, manuscript.
Benigno, P., and Woodford, M. (2003). Optimal monetary and fiscal policy: A linear-quadratic
approach. NBER macroeconomics annual, 18, 271-333.
Bjørnland, H. C., and Thorsrud, L. A. (2015). Commodity prices and fiscal policy design:
Procyclical despite a rule. CAMP Working Paper Series No 5. Centre for Applied Macro
and Petroleum Economics.
45
Blundell, R., and Bond, S. (1998). Initial Conditions and Moment Restrictions in Dynamic Panel
Data Models. Journal of Econometrics, 87, 115-143.
Calderón, C., and Nguyen, H. (2016). The Cyclical Nature of Fiscal Policy in Sub-Saharan Africa.
Journal of African Economies, 25(4), 548-579.
Calderón, C., and Schmidt‐Hebbel, K. (2003). Macroeconomic Policies and Performance in
Latin America. Journal of International Money and Finance, 22(7), 895-923.
Calderón, C., and Schmidt-Hebbel, K. (2008a). The Choice of Fiscal Regime in the World.
Working paper No. 487, Central Bank of Chile.
Calderón, C., and Schmidt-Hebbel, K. (2008b). What drives the Choice of Inflation Targets in
the World. Unpublished manuscript, Central Bank of Chile.
Calderón, C., and Schmidt-Hebbel, K. (2008c). Business Cycles and Fiscal Policies: The Role of
Institutions and Financial Markets, Working paper No. 481, Central Bank of Chile.
Calderón, C., Duncan, R., and Schmidt-Hebbel, K. (2016). Do Good Institutions Promote
Countercyclical Macroeconomic Policies? Oxford Bulletin of Economics and Statistics,
78(5), 650-670.
Castro, V. (2011). The impact of the European Union fiscal rules on economic growth. Journal
of Macroeconomics, 33(2), 313–26.
Céspedes, L., and Velasco, A. (2012). Macroeconomic Performance During Commodity Price
Booms and Busts. IMF Economic Review, 60, 570-599.
Chinn, M., and Ito, H. (2008). A new measure of financial openness. Journal of Comparative
Policy Analysis, 10(3), 307–20.
Cruz, C., Keefer, P., and Scartascini, C. (2016). Database of Political Institutions Codebook,
2015 Update (DPI2015). Inter-American Development Bank.
Dabla-Norris, M. E., Gollwitzer, S., Kvintradze, E., Prakash, M. T., Zanna, L. F., Allen, R., and
Lledo, V. D. (2010). Budget institutions and fiscal performance in low-income countries.
IMF Working Paper WP/10/80.
Debrun, X., and Kumar, M. S. (2007). The Discipline-Enhancing Role of Fiscal Institutions:
Theory and Empirical Evidence. IMF Working Paper WP/07/171.
Debrun, X., Kinda, T., Curristine, T., Eyraud, L., Harris, J., and Seiwald, J. (2013). The Functions
and Impact of Fiscal Councils. IMF Policy Paper; July. International Monetary Fund.
Easterly, W., and Kraay, A. (2000). Small States, Small Problems? Income, Growth, and
Volatility in Small States. World Development, 28(11), 2013–2027.
Elbadawi, I., Schmidt-Hebbel, K., and Soto, R. (2015). Why do countries have fiscal rules? in
Economic Policy in Emerging-Market Economies, R. Caballero and K. Schmidt-Hebel, eds.,
Santiago: Central Bank of Chile.
Elbadawi, I., Soto, R., and Youssef, H. (2017). Sovereign Wealth Funds and Macroeconomic
Stabilization in the Home Economy. Mimeo, Economic Research Forum.
46
Escolano, M. J., Eyraud, L., Tuladhar, M. A., Badia, M. M., and Sarnes, M. J. (2012). Fiscal
performance, institutional design and decentralization in European Union countries.
IMF Working Paper WP/12/45.
Eslava, M. (2011). The political economy of fiscal deficits: a survey. Journal of Economic
Surveys, 25(4), 645-673.
Eyraud, L., and Lusinyan, L. (2013). Vertical fiscal imbalances and fiscal performance in
advanced economies. Journal of Monetary Economics, 60(5), 571-587.
Favaro, E. (2008). Small States, Smart Solution. Improving Connectivity and Increasing the
Effectiveness of Public Services, E. Favaro (ed). The World Bank, Washington, DC.
Feld, L. P., and Schnellenbach, J. (2010). Fiscal Federalism and Long-Run Macroeconomic
Performance. International Studies Program Working Paper 10/09, Andrew Young
School of Policy Studies, Georgia State University.
Frankel, J. A., Vegh, C. A., and Vuletin, G. (2011). On Graduation from Fiscal Procyclicality.
NBER Working Paper No. 17619.
Frenkel, J., Goldstein, M., and Masson, P. (1991). Characteristics of a Successful Exchange Rate
System. IMF Occasional Paper No. 82.
Ghosh, A. R., Kim, J. I., Mendoza, E. G., Ostry, J. D., and Qureshi, M. S. (2013). Fiscal fatigue, fiscal
space and debt sustainability in advanced economies. The Economic Journal, 123(566).
Giavazzi, F., and Pagano, M. (1988). The advantage of tying one’s hands: EMS discipline and
central bank credibility. European Economic Review, 32(5), 1055–75.
Halac, M., and Yared, P. (2014). Fiscal rules and discretion under persistent
shocks. Econometrica, 82(5), 1557-1614.
Heckman, J. J. (1978). Dummy Endogenous Variables in a Simultaneous Equations System.
Econometrica, 46, 931–960.
Hodrick, R. J., and Prescott, E. C. (1997). Postwar U.S. Business Cycles: An Empirical
Investigation. Journal of Money, Credit, and Banking, 29(1), 1–16.
Hughes Hallett, A., and Lewis, J. (2005). Fiscal Discipline before and after EMU - Permanent
Weight Loss or Crash Diet? Working Papers 05-W16, Vanderbilt University Department
of Economics.
Ilzetzki, E., and Vegh, C. A. (2008). Pro-Cyclical Fiscal Policy in Developing Countries: Truth or
Fiction? NBER Working Paper No. 14191.
IMF. (2006). Inflation Targeting and the IMF. Mimeo, prepared by Monetary and Financial
Systems Department, Policy and Development Review Department, and Research
Department.
IMF. (2013). The Functions and Impact of Fiscal Councils. IMF Policy Paper, (Washington, DC:
International Monetary Fund).
47
Jordà, Ò., and Taylor, A. M. (2016). The Time for Austerity: Estimating the Average Treatment
Effect of Fiscal Policy. Economic Journal, 126, 219–255.
Klein, M. W., and Shambaugh, J. C. (2008). The dynamics of exchange rate regimes: fixes, floats,
and flips. Journal of International Economics, 75(1), 70–92.
Leith, C., and Wren-Lewis, S. (2005). Fiscal stabilization policy and fiscal institutions. Oxford
Review of Economic Policy, 21(4), 584-597.
Lucotte, Y. (2012). Adoption of inflation targeting and tax revenue performance in emerging
market economies: An empirical investigation. Economic Systems, 36(4), 609-628.
Manzano, O., and Rigobon, R. (2007). Resource curse or debt overhang, in Natural Resources:
Neither Curse nor Destiny, D. Lederman, W.F. Maloney (Eds.), The World Bank,
Washington, D.C.
Milesi-Ferrett, G. M., and Razin, A. (1996). Current Account Sustainability. Princeton Studies in
International Finance, 81.
Minea, A., and Tapsoba, R. (2014). Does inflation targeting improve fiscal discipline? Journal of
International Money and Finance, 40, 185–203.
Minea, A., and Villieu, P. (2009). Threshold effects in monetary and fiscal policies in a growth
model: Assessing the importance of the financial system. Journal of Macroeconomics,
31(2), 304-319.
Nerlich, C., and Reuter, W. H. (2015). Fiscal rules, fiscal space and procyclical fiscal policy. ECB
Working Paper No. 1872.
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica, 49, 1417–1426.
Polity IV Project. (2010). Political Regime Characteristics and Transitions, 1800–2010
Database, www.systemicpeace.org/polity/polity4.htm
Reinhart, C., and Rogoff, K. S. (2004). The modern history of exchange rate arrangements: a
reinterpretation. The Quarterly Journal of Economics, 119(1), 1–48.
Riascos, A., and Vegh, C. A. (2004). Procyclical Fiscal Policy in Developing Countries: The Role
of Capital Market Imperfections. University of California at Los Angeles, manuscript.
Rogoff, K. (1990). Equilibrium Political Budget Cycles. American Economic Review, 80, 21-36.
Schaechter, A., Kinda, T., Budina, N., and Weber, A. (2012). Fiscal Rules in Response to the
Crisis—Towards the “Next-Generation” Rules. A New Dataset. IMF Working Paper
WP/12/187.
Schmidt-Hebbel, K. (2016). Fiscal Institutions in Resource-Rich Economies: Lessons from
Chile and Norway. In I. Elbadawi and H. Selim, eds., Understanding and Avoiding the Oil
Curse in Resource-rich Arab Economies. Cambridge: Cambridge University Press.
Schmidt-Hebbel, K., and Soto, R. (2017). Fiscal Rules in the World, in Rethinking Fiscal Policy
After the Crisis, Ľudovít Ódor (ed.), Cambridge University Press.
48
Soto, R. (2016). Resource abundance and dependency and the quality of economic growth.
Mimeo, Pontificia Universidad Católica de Chile.
Talvi, E., and Vegh, C. (2005). Tax base variability and procyclical fiscal policy in developing
countries. Journal of Development Economics, 78(1), 156–90.
49
Appendices
Appendix Table 1 Country List
Albania El Salvador Kuwait Qatar
Algeria Estonia Latvia Romania
Argentina Ethiopia Lebanon Russian Federation
Armenia Finland Libya Saudi Arabia
Australia France Lithuania Senegal
Austria Gabon Luxembourg Serbia
Azerbaijan Gambia, The Madagascar Seychelles
Bangladesh Germany Malawi Sierra Leone
Belarus Ghana Malaysia Slovak Republic
Belgium Greece Mali Slovenia
Bolivia Guatemala Mexico South Africa
Botswana Guinea Moldova Spain
Brazil Guinea-Bissau Morocco Sri Lanka
Bulgaria Haiti Mozambique Suriname
Burkina Faso Honduras Namibia Sweden
Cameroon Hong Kong Netherlands Switzerland
Canada Hungary New Zealand Syria
Chile Iceland Nicaragua Tanzania
China India Niger Thailand
Colombia Indonesia Nigeria Togo
Congo, Dem. Rep. Iran, Islamic Rep. Norway Trinidad & Tobago
Congo, Rep. Ireland Oman Tunisia
Costa Rica Israel Pakistan Turkey
Cote d'Ivoire Italy Panama Uganda
Croatia Jamaica Papua New Guinea Ukraine
Cyprus Japan Paraguay United Kingdom
Czech Republic Jordan Peru United States
Denmark Kazakhstan Philippines Uruguay
Dominican Republic Kenya Poland Venezuela
Ecuador Korea, Rep. Portugal Vietnam
Egypt Yemen
Appendix Table 2 Definitions and Sources of Variables used in Regression Analysis
Variable Definition and construction Source
Dependency Ratio
Population between 15 and 64 years of age as
share of total population
World Bank World Development
Indicators
50
Exchange Rate Regime Dummy variable taking value 1 if the country
is dollarized, has a currency board, or belong s
to a monetary union. Any other system is not
considered as a fixed regime.
Reinhart and Rogoff (2004) de jure
classification, extended using IMF
country reports.
Export Concentration Herfindahl’s index of 26 exported goods
categories, measured annually in US dollars,
using data from UN COMTRADE database.
Soto (2016)
Federalism Dummy variable = 1 if the country defines
itself formally as a federal entity.
Information from Forum of
Federations web page.
Financial Development Domestic credit to private sector (% of GDP World Development Indicators by
the World Bank
Financial Openness Index based on the binary dummy variables
that codify the tabulation of restrictions on
cross-border financial transactions reported in
the IMF's Annual Report on Exchange
Arrangements and Exchange Restrictions
Chinn and Ito database (2014).
Fiscal Balance General Government primary balance Databases of the IMF, World Bank,
ECLAC, OECD, Asian Development
Bank and African Development
Bank
Fiscal Rules A fiscal rule is defined as a permanent
constraint on fiscal policy through simple
numerical limits on budgetary aggregates.
IMF Fiscal Rules Dataset, 1985-
2015.
Government Gross Debt All liabilities that require payments of interest
and/or principal by the debtor to the creditor
at a date or dates in the future.
IMF World Economic Outlook
Database
Government Revenue
Instability
Three-year rolling coefficient of variation of
real government revenue (including taxes,
social contributions, grants receivable, and
other revenue).
Databases of the IMF, World Bank,
ECLAC, OECD, Asian Development
Bank and African Development
Bank
Government Stability Assessment of a government’s ability to carry
out its declared program and stay in office.
The risk rating is the sum of three
subcomponents (government unity, legislative
strength and popular support).
International Country Risk Guide
database, from PRS Group.
Inflation Measured by the log change of the consumer
price index or GDP deflator.
World Development Indicators by
the World Bank
Inflation Targeting Dummy variable: 1 if the central bank operates
formally an inflation targeting scheme, and 0
otherwise.
Central Bank News webpage and
database.
Monetary Union Binary variable that takes value 1 if the
country formally belongs to a monetary union
and 0 otherwise.
Own elaboration.
Natural Resource Rents Rents are the difference between the value of
production for a stock of minerals at world
prices and their total costs of production.
World Development Indicators by
the World Bank
Political Accountability Checks and balances indicators Cruz et al. (2016)
Political Participation Democracy and Polity2 indices of the Polity IV
database
Developed by Integrated Network
for Societal Conflict Research
(INSCR).
Procyclicality of Fiscal
Balance
Ten-year rolling correlation of HP-filtered
fiscal balance to HP-filtered GDP (both in
constant 2000 US$).
Databases of the IMF, World Bank,
ECLAC, OECD, Asian Development
Bank and African Development
Bank
51
Procyclicality of Government
Expenditures
Ten-year rolling correlation of HP-filtered
expenditures of general government to HP-
filtered GDP (both in constant 2000 US$).
Expenditure consists of total expense and the
net acquisition of nonfinancial assets.
IMF Government Finance Statistics
and World Economic Outlook
Databases.
Real per capita GDP Ratio of total GDP to total population. GDP is in
1985 PPP-adjusted US$. Growth rates are
obtained from constant 1995 US$ per capita
GDP series.
World Development Indicators by
the World Bank
Remittances Workers' remittances and compensation of
employees, received.
World Development Indicators by
the World Bank
Sovereign Wealth Funds Dummy variable taking value 1 if the country
has an international sovereign wealth fund,
and 0 otherwise.
Own elaboration.
Terms-of-trade shocks Log difference of the terms of trade. Terms of
trade are defined as customary.
World Development Indicators by
the World Bank
Trade Volume Exports + Imports (as share of GDP) World Development Indicators by
the World Bank
52
Appendix Table 3 Stage 2: Linear Regression Models for Fiscal Rules
------------------------------------------------------------------------------------------------------------ National Rules --------------------------------------------------------------------------- Any Rule Any Expenditure Revenue Budget Balance Debt ------------------------------------------------------------------------------------------------------------ Political 0.0153*** 0.00539** 0.00111 0.000972 0.000172 0.000513 Participation (5.30) (2.00) (0.58) (0.79) (0.08) (0.26) Federalism 0.0105 0.0799*** 0.00295 -0.00755 0.134*** 0.128*** (0.45) (3.77) (0.19) (-0.78) (7.14) (8.48) Checks and 0.0192*** 0.00623 0.0104*** 0.00690*** -0.000743 -0.00235 Balances (3.90) (1.36) (3.14) (3.30) (-0.19) (-0.71) Government -0.00137 0.00356 -0.000610 0.00148 -0.00299 0.00294 Stability (-0.37) (1.04) (-0.25) (0.94) (-1.04) (1.19) Monetary Union 0.378*** 0.0256 0.0312* 0.0627*** -0.0377** 0.00388 (13.13) (1.11) (1.89) (6.17) (-2.03) (0.24) Fixed Exchange 0.112*** -0.0318* 0.00772 -0.0142* -0.0647*** 0.00634 Regime (5.67) (-1.72) (0.58) (-1.70) (-4.21) (0.48) Inflation Target 0.192*** 0.194*** 0.0764*** -0.00119 0.158*** 0.126*** (8.40) (9.09) (5.09) (-0.13) (9.06) (8.35) Financial 0.361*** 0.302*** 0.0907*** 0.0426*** 0.218*** 0.0639*** Openness (14.21) (12.65) (5.17) (3.96) (11.03) (3.77) Financial -0.0192* 0.0176 -0.0249*** 0.0159*** 0.0183** 0.0271*** Development (-1.67) (1.63) (-3.22) (3.24) (2.05) (3.51) Economic 0.0201** -0.00229 0.0374*** 0.00469 -0.0268*** -0.0511*** Development (1.97) (-0.24) (5.39) (1.09) (-3.41) (-7.51) Dependency 0.248** 0.134 -0.129* -0.100** 0.202** 0.481*** Ratio (2.26) (1.34) (-1.81) (-2.22) (2.45) (6.51) Resources -0.0500*** -0.0313*** -0.0247*** -0.00249 -0.00261 0.0234*** Rents (-5.99) (-4.75) (-5.49) (-0.92) (-0.53) (5.56) Workers -0.000497 -0.0195*** -0.00537* 0.000757 -0.0147*** 0.000597 Remittances (-0.11) (-4.69) (-1.81) (0.40) (-4.28) (0.20) Any rule -0.0661** (instr.) (-2.10) Any national -0.0685** Rule (instr.) (-1.98) Exp. Rule 0.181*** (instr.) (3.71) Revenue Rule 36.65*** (instr.) (13.42) Budget Balance -0.0911** Rule (instr.) (-2.08) Debt Rule -0.260*** (instr.) (-2.82) Constant -1.982*** -1.009*** -0.150 0.316* -0.627** -1.251*** (-4.55) (-2.67) (-0.57) (1.91) (-2.06) (-4.59) ------------------------------------------------------------------------------------------------------------ Observations 2744 2744 2744 2744 2744 2744 Number of Countries 118 118 118 118 118 118 ------------------------------------------------------------------------------------------------------------ t statistics in parentheses * p<0.1, ** p<0.05, *** p<0.01
53
Appendix 1 Generating Instruments for Sovereign Wealth Funds
Following the results of Elbadawi et al. (2017) we first estimate a probit model for a
binary variable taking value 1 if country “I” has a sovereign wealth fund in place at time “t”
and zero otherwise. The literature suggests the existence of several potential determinants of
the likelihood of accumulating net foreign assets in the form of sovereign wealth funds (Beck
and Fidora, 2008).
The first area relates to export proceeds. Earlier commodity price booms illustrate the
adverse effect on competitiveness of large real appreciations induced by using these windfall
gains for domestic expenditures, particularly when the gains are transitory (Céspedes and
Velasco, 2012). Consequently, some countries have sought to deal with these concerns by
saving a share of the gains in SWFs. We thus control for the level of foreign trade (as share of
total GDP), export concentration (measured by the Herfindhal index, as computed by Soto
2016), and the size of natural resource rents (from the World Bank Database).
The second area of relates to macroeconomic stance and general development level.
More developed countries, on average, tend to have better macroeconomic management
which, more often than not, include policies of accumulating foreign reserves. Diversification
of these reserves into potentially higher-yielding assets have usually entailed transferring
them from the control of the central bank to the Treasury or to quasi-public entities, such as
SWFs, with the mandate to pursue financial strategies aiming at higher long-run returns
(Aizenman and Glick, 2009). We control for development levels using real per capita GDP in
US$. We also control more directly for macroeconomic mismanagement using the inflation
rate as a measure of price instability.
A third group of factors behind the growth of SWFs relates to government revenue
structures and, particularly, tax revenue instability. A large number of SWFs had been set up
to insulate the budget and economy from commodity price volatility and external shocks.
Their investment horizons and liquidity objectives resemble central banks reserve managers,
in view of their role in countercyclical fiscal policies to smooth boom/bust cycles. We control
for government revenue instability using the coefficient of variation of tax revenues computed
as a rolling three-year. We also control for the fact that countries with federal structures may
find it more appealing –and perhaps more challenging—to set up a SWF as a way to increase
independence of the federal government from the subnational units. A dummy variable is
used in the empirical analysis.
Finally, we also control for idiosyncratic factors in the GCC economies. Gulf economies
have not only allocated a significant fraction of their wealth in SWFs but have also been the
pioneers in their implementation, following Kuwait’s initiative of 1953. These resource-rich
economies are subject to the highly volatile oil prices; SWFs partly serve the purpose of
stabilizing government revenues which would otherwise mirror the volatility of oil and
commodity prices (Barnett and Ossowski, 2002). We use a dummy to identify GCC countries.
54
Column 1 in Appendix Table 4 reports the results of estimating probit regressions of
the incidence of FRs for 146 countries using annual data in the period 1984-2015.9 As
reported, more open countries that enjoy higher levels of resource rents and have
concentrated exports tend to have a higher probability of implementing SWFs. Likewise, more
developed economies –which usually also have lower levels of price instability—are more
likely to have established sovereign wealth funds. Finally, the evidence confirms that
instability in tax revenues is an important reason for governments to implement SWFs and
the preference of the oil-rich economies of the GCC for this type of fiscal institution. We use
these results to generate a prediction of the likelihood of having a SWF in place for each year
of the period 1984-2015, conditional on the observed values of fundamentals. This predicted
variable is a consistent instrument for SWF but is not efficient. Efficiency obtains when the
SWF dummy is filtered against the fundamentals and the instrument obtained in the previous
step. Results are contained in Panel B of Table 1.
Appendix Table 4
Instrumenting Sovereign Wealth Funds
Sovereign Wealth Fund
1st stage
Sovereign Wealth Fund
2nd stage
Economic Development 2.827*** 0.046***
Resource Rents 2.183*** 0.026***
Price Instability -0.057*** 0.000
Trade Volume 1.707*** 0.158***
Exports Concentration 0.926** 0.087***
Revenue Instability 3.865** 0.102*
Federalism -3.855*** 0.231***
GCC country 5.510*** -0.011
Constant -103.050*** -2.533***
First Stage 𝑺𝑾�̂� - 0.157***
Observations 3135 3135
Countries 146 146
Note: * p<0.1, ** p<0.05, *** p<0.01
9 We also controlled for political factors and national governance using data from ICRG but the results proved to be statistically and economically insignificant.
55
Contents
1 Introduction ................................................................................................................................................................ 1
2 Fiscal Rules and Fiscal Performance: Measures, Interactions, and Policy Optimality ................. 3
2.1 Interactions between Fiscal Performance, Fiscal Institutions, Economic Conditions, and
Political-Economy Features ................................................................................................................................. 3
2.2 Measures of Fiscal Performance ................................................................................................................ 3
2.3 Types of Fiscal Rules ....................................................................................................................................... 5
2.4 Optimal Fiscal Policy ....................................................................................................................................... 7
2.5 A simple optimal fiscal policy rule for an open economy ............................................................... 8
2.6 Relations between types of fiscal target rules and fiscal performance measures ............. 10
3 Previous Empirical Evidence on Fiscal Rules and Fiscal Performance .......................................... 12
3.1 Fiscal Performance ....................................................................................................................................... 12
3.2 Fiscal Rules and Fiscal Performance ..................................................................................................... 14
4 Are Small Economies Different? ..................................................................................................................... 15
5 Empirical Methodology ...................................................................................................................................... 16
5.1 Fiscal Rules as Treatments ....................................................................................................................... 17
5.2 On the Endogeneity of Fiscal Rules ....................................................................................................... 19
6 Empirical Analysis I: Instrumenting Fiscal Rules .................................................................................... 20
7 Empirical Analysis II: Assessing the Impact of Fiscal Rules ................................................................ 26
7.1 Measures of fiscal performance ............................................................................................................... 26
7.2 Determinants of Fiscal Performance ..................................................................................................... 27
7.3 Main Results on Procyclicality of Government Expenditures ..................................................... 29
7.4 Main Results on Procyclicality of Fiscal Balance .............................................................................. 32
7.5 Main Results on Fiscal Balances .............................................................................................................. 34
7.6 Main Results on Government Debt ......................................................................................................... 37
8 De Facto and de Jure Rules ................................................................................................................................ 39
9 Conclusions .............................................................................................................................................................. 41
References .................................................................................................................................................................... 44
Appendices ................................................................................................................................................................... 49