What do OECD countries cut first when faced with fiscal adjustments?
Sanz, Ismael
1. Introduction
As OECD member countries have dramatically worsened their public
finances, they will be forced to undertake budgetary cuts in the next
years. Developed countries will increase their public deficit to 8.8% of
GDP in 2010, compared with the 2.1% (on average) of GDP for the period
extending between 2000 and 2007, whereas the public debt will increase
to 100.2% of GDP, up from 72.6% during the 2000-2007 period (OECD
Economic Outlook 85, June 2009). Fiscal discipline will require cuts in
government expenditure, leading to a trade-off between different
components of government expenditure that will affect the composition of
government expenditure. In this article we explore the relationship
between components of government expenditure and government size during
the 1970-2007 period for a sample of 25 OECD countries to shed light on
how fiscal discipline might influence public spending composition in the
future.
Many authors (Dunne, Pashardes, and Smith 1984; Borge and Rattso
1995; Sturm 1998; Tridimas 2001; Shelton 2007) have underscored the
surprisingly little research devoted to the determinants of the
composition of government expenditure. Moreover, Baqir (2002) claims
that most of the studies analyzing the effects of aggregate government
expenditure on its composition have focused on the economic
classification of government spending. Furthermore, those few studies
examining the functional disaggregation of government expenditure have
concentrated on particular functions--primarily social expenditure,
including education and health (Baqir 2002), and occasionally social
welfare expenditure (Ravallion 2002)--or on the composition of local
government spending (Borge and Rattso 1995). The contribution of this
article is to use the functional disaggregation of consolidated
government expenditure, Classification of the Functions of Government
(COFOG). To our knowledge, this is the first study on the effects of
fiscal consolidation on the composition of government expenditure by
functions for the OECD.
The assessment of the effects of aggregate government expenditure
is of great interest in the future, especially in the context of
budgetary cuts and cost controls. Fiscal consolidation affects economic
growth through its impact on the composition of public expenditure.
Along these lines, some endogenous growth models incorporate the
composition of government spending that is capable of yielding
steady-state effects (Devarajan, Swaroop, and Zou 1996; Gemmell,
Kneller, and Sanz 2009). Moreover, Davoodi and Zou (1998) show that
there is an optimal composition of government expenditures in which the
optimal share of each component equals its growth elasticity, relative
to all of the growth elasticities. Therefore, by changing the
composition of government expenditure, fiscal consolidation can approach
or deviate the structure of public spending from its optimal structure.
In order to investigate this aspect, section 2 reviews the existing
literature on the effects of fiscal consolidation on the composition of
government expenditures. In section 3, we introduce the data to be used
and the empirical methodology. In section 4, we analyze how the
composition of government expenditure changes when the public sector
size decreases in a dynamic model framework. Section 5 checks the
robustness of our conclusions by investigating the impact on the
composition of government spending decreases in the public debt and
deficit. In section 6, we draw the most significant conclusions.
2. Fiscal Consolidation and the Composition of Government
Expenditure
During the period ranging from 1970 to 2007, the share of
government expenditure in GDP has increased in the OECD, from 30.5% in
1970 to 42.2% in 2007 (OECD: National Accounts. Volume IV: General
Government Accounts). However, this expansion has fluctuated over the
course of the four decades. During the 1970s and the early years of the
1980s the public sector increased its size constantly. This trend was
interrupted in 1983, when public expenditure as a share of GDP became
stable. At the beginning of the 1990s public expenditure began to
increase its size again until 1993, the peak for the whole period.
Thereafter, OECD countries have increasingly implemented government
spending reforms aimed toward more controlled government spending and
active deficit management (Tanzi and Schuknecht 2000).
The reduction of the public sector size does not necessarily lead
to proportional decreases in the components of government spending, but
it may change the composition of government expenditures by particularly
affecting some of its components while protecting others. Several
studies have analyzed the effects of fiscal consolidation on the
composition of government expenditures, focusing on two specific
components: public investments and social spending (including education,
health, and social welfare expenditure). These studies predict two
different and opposite outcomes. The first strand of studies claims that
fiscal adjustments will affect investments while protecting social
spending. Thus, Roubini and Sachs (1989) claim that during a time of
fiscal consolidation, public investments are the first to be reduced
because these represent the least rigid component of expenditures. Oxley
and Martin (1991) also contend that political reasons make it easier to
diminish or postpone investment spending than current expenditure.
Furthermore, Sturm (1998) suggests that myopic governments in need of
budgetary cuts reduce those less visible and long-term expenditures in
order to minimize the political costs associated with government
spending cutbacks. Gomes and Pouget (2008) elaborate a model in which
international tax competition drives tax rates down, reducing the
externality of public capital and thereby leading governments to
decrease public investment. De Mello (2008) argues that current spending
is increasingly downward rigid, and therefore, fiscal adjustments
compress public investment. Finally, Easterly, Irwin, and Servrn (2008)
argue that if governments reduce productive spending they are improving
short-term cash deficit, but they might be worsening fiscal imbalances
in the long term if the foregone growth reduces the present value of
future government revenues by more than the immediate improvement in the
cash deficit. Nevertheless, these authors contend that too much emphasis
on short-term cash flows leads governments to reduce productive spending
at times of fiscal adjustment and to protect non-productive spending.
Along these lines, Henrekson (1988; for Sweden over the period from 1950
to 1984); Sturm (1998; for a sample of 22 OECD countries over the period
from 1980 to 1992); Jonakin and Stephens (1999; for a sample of five
Central American countries over the period from 1975 to 1993); Mahdavi
(2004; for 47 developing countries over the period from 1972 to 2001);
and Akitoby et al. (2006; for a sample of 51 developing countries over
the period from 1970 to 2002) find that fiscal adjustments particularly
affect public investments.
Therefore, we could expect fiscal consolidation to fall primarily
on public investments and to protect the rest of the expenditures. In
fact, studies examining the effects of fiscal adjustments on pro-poor
expenditures---mainly social expenditures--predict that budgetary cuts
will not primarily affect social spending. Thus, Ravallion (1999) claims
that if cutting expenditures save taxes to the non-poor, these voters,
in turn, would be more willing to protect pro-poor expenditure.
Furthermore, Ravallion (2000) contends that poor groups could build
influential interest groups with Non-Governmental Organizations or
non-poor groups interested in avoiding the external costs of poverty.
Accordingly, Snyder and Yackovlev (2000; for a sample of 19 Latin
American and Caribbean countries during the period from 1970 to 1996)
and Cashin et al. and Baqir (2001 and 2002; for a sample of 179 and 167
countries during the period from 1985 to 1998, respectively) find that
education and health expenditure are isolated from fiscal adjustments.
A second strand of studies predicts that budgetary cuts will affect
social expenditures and protect productive expenditures. Aubin et al.
(1988) argue that reducing investments---a type of productive
expenditure--has more adverse political effects than does decreasing
public consumption and wages because the former is more visible. In
fact, Alesina, Perotti, and Tavares (1998) show that if anything,
adjustments, primarily based on public transfers and wages, increased
the probability of survival of governments over the period from 1960 to
1995 in a sample of 19 OECD countries. Moreover, Tanzi (2000) maintains
that globalization will reduce government revenues and expenditure
because of the tax competition among jurisdictions and increased
mobility of factors. These authors suggest that the reduction of
government spending will not be proportional but will affect social
spending in particular and preserve productive expenditures that enhance
countries' competitiveness and attractiveness to foreign direct
investment (FDI). Along these lines, Keen and Marchand (1997) elaborate
a model in which governments encourage country competitiveness by
raising the allocation to productive expenditures above its optimal
level and contracting utility-enhancing spending, such as social welfare
spending. Furthermore, Ghate and Zak (2002) elaborate a model in which
politicians maximize votes, whereas voters support politicians depending
on the transfers they receive and the output growth. As a consequence,
at a first stage social welfare expenditure drives the growth of
government, but then a threshold emerges at which, in order to maintain
positive output growth, governments reduce aggregate government
expenditure by cutting social welfare expenditure. Finally, Drazen and
Eslava (2010) elaborate a model of the political budget cycle in which
incumbents try to influence voters by shifting the composition of
government spending toward functions that are particularly attractive to
voters. These authors find evidence that voters favor infrastructure and
education spending because these two functions increase in electoral
years at the cost of transfers for a sample of all municipalities in
Colombia over the period 1987 to 2002. Accordingly, Tanzi and Schuknecht
(1997) show that industrialized countries that undertook reforms in the
size of the public sector in the mid1980s, such as New Zealand and
Chile, accomplished it through the reduction of public subsidies and
transfers. Furthermore, Ravallion (2002) finds that decreases in
aggregate government expenditure in the decade of the 1980s and the
1990s in Argentina have led to more-than-proportional cuts in education,
health, housing, and social security spending, whereas increases in
aggregate government spending did not increase the size of social
spending at all.
3. Data and Methodology
Data for government expenditures is built on OECD publication
National Accounts. Volume IV: General Government Accounts. From this
source we elaborate time series of government spending by functions over
the period extending from 1995 to 2007. We use the standard COFOG
(United Nations 2000), which considers the following 11 components of
government expenditures (in order of average magnitude in the OECD):
social security; education; health; general public services; economic
affairs (distinguishing between transport and communication and other
economic affairs); defense; public order and safety; housing and
community amenities; recreation, culture, and religion; and environment
protection. Data for the 1970-2000 period are built on National
Accounts: Volume II: Detailed Tables, which follows the previous COFOG
(United Nations 1981). There are minor differences between the previous
and the updated COFOG classifications (IMF 2002), mainly the inclusion
of the new category 'environment protection,' which was
previously aggregated with housing and community amenities, and the
removal of the component 'other non-classified functions'
(mainly interest payments), which is now aggregated with general public
services. Nevertheless, there is a subcategory among the new general
public services named public debt transactions, which mainly corresponds
to the former other non-classified functions.
We have used the time series over the period 1995 to 2007 based on
the new COFOG classification and extended it back to 1970 employing the
previous COFOG. (1) Table 1 illustrates the correspondence between the
previous and the new COFOG classifications. We end up with time series
for 10 components of government spending (summing general public
services and public order and safety under the aggregated public
services) by functions over the 1970-2007 period for 25 OECD countries,
all developed countries, excluding the Central and Eastern European
Countries (Poland, Czech Republic, Hungary, and Slovakia), for which
there are no data available before 1990, and Turkey, with no information
after 2001. (2)
Table 1 also shows the functions that, according to Kneller,
Bleaney, and Gemmell (1999), have a positive economic growth effect and
are hence considered as productive: education, health, defense, public
order and safety, transport and communication, housing and community
amenities, environment protection, and general public services
(excluding public debt transactions). In the last column, we further
aggregate functions between those with a social service character (that
is, functions that enhance utility) and those without such social
character. Among those functions with a social service character,
previous literature has included social protection, education, and
health (Ravallion 2002).
[FIGURE 1 OMITTED]
In sum, there are functions of government spending that are mainly
utility-enhancing (social security), growth-enhancing (housing,
environment, defense, transport and communication, public order and
safety, and general public services), both (education and health), and
those that are neither utility nor growth-enhancing (economic affairs
and recreation). (3) The first strand of the literature discussed in the
previous section claims that fiscal adjustment will protect social
spending at the expense of productive but non-social spending. The
second strand contends that budgetary cuts disproportionally fall over
social but non-productive spending, shielding productive spending.
Therefore, the two hypotheses concur in predicting that education and
health would be protected from fiscal adjustments and that economic
affairs and recreation, culture, and religion would be disproportionally
reduced in fiscal consolidations. For the rest of the functions that are
considered social but non-productive (social protection) or productive
but non-social (transport and communications, defense, public order and
safety, general public services, housing and environment protection),
the two hypotheses predict contradicting effects from budgetary cuts.
In the empirical analysis we do not consider public debt
transactions because this spending is exogenously determined, and it is
the consequence of the fiscal imbalances that OECD countries will have
to reduce in the next years. We analyze how the public sector size
affects the share of each function in the aggregate government spending,
excluding these interest payments.
Figure 1 shows the evolution of aggregate government spending
excluding interest payment (percentage of GDP) in the OECD and the share
devoted to the three most important functions: social welfare,
education, and health (as percentage of total government spending
excluding interest payments). Social security shows a similar pattern
compared with aggregate government spending. From 1970 to 1993, total
government spending as a share of GDP increased 13 percentage points,
from 29.6% of the GDP to 42.6% of the GDP. More than half of this
increase is explained by social security spending, which in 1994 reached
its peak, accruing 37.9% of all public spending. After 1995, both the
public sector size and social security have decreased, although the
latter to a lesser extent. In contrast, education and health show a
divergent pattern compared with the public sector size. These
expenditures decrease their share in total government expenditure from
1970 to 1986, from 26.4% to 24.9%, just when the public sector size was
growing more heavily. From 1986 to the mid-1990s, education and health
maintained an almost constant share in aggregate government spending.
These expenditures started to rise later on, from 1994 to 2007, the same
period during which the public sector size diminished. Figure 2 confirms
the same pattern for the broader definitions of non-productive spending
and productive spending. The share of non-productive spending follows
the same evolution as that of aggregate government spending: It has
increased until the mid-1990s and has slightly decreased since then.
Productive spending has obviously followed the opposite pattern, showing
a negative correlation with aggregate government spending.
[FIGURE 2 OMITTED]
Our analysis uses annual panel data and the Pooled Mean Group (PMG)
methodology proposed by Pesaran, Shin, and Smith (1999). These authors
show that the assumption of homogeneity of the short-run parameter
estimates across countries cannot be accepted and that this assumption
may be a more serious problem than the endogeneity bias generated by the
inclusion of lagged dependent variables and can lead to inconsistent and
misleading results even for large T and large N. To overcome this bias
they suggest the use of the PMG estimator. This allows heterogeneous
constants and marginal short-run effects across countries to be
accommodated, while maintaining homogeneity of the long-run responses.
The major advantage of this approach is that it makes full use of the
available time-series information and provides estimates of both
long-run and short-run parameters. We introduce dynamics into the model
to capture the fact that some of the government expenditures show a high
degree of rigidity since some of the expenditures are previously
committed, as are those related to social security and public
employees' wages. Moreover, the incrementalist decision-making
literature (Borge and Rattso 1995) has shown the decisive advantage of
the status quo in the determination of government expenditure. We
estimate the following PMG estimation for each of the 10 components of
government expenditure (aggregating general public services and public
order and safety) and inferring coefficients for environment protection
(the smallest function) from the budget restriction:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where i denotes the country, t is time, g is the log share of each
of the 10 functions in the aggregate government spending, F is a matrix
including the log of the public sector size and control variables, and
[[epsilon].sub.i,t] is a classical error term. The parameter vectors
[[phi].sub.i] and [beta], respectively, capture the error correction and
(homogeneous) long-run growth effects, while [[lambda].sub.i,m] and
[[gamma].sub.i,k] capture the heterogeneous short-run responses to g and
F, respectively (with lag lengths M, K = 1).
Among control variables we introduce variables based on the
previous literature (Sanz and Velazquez 2007): per capita income, total
population, and young and elderly shares in total population. (4) We
also include openness since there are two hypotheses (the compensation
and the efficiency hypothesis) that propose that globalization affects
the composition of public expenditures (Gemmell, Kneller, and Sanz
2008). The compensation hypothesis contends that the economic insecurity
introduced by globalization leads to expanding social expenditures,
whereas under the efficiency hypothesis, globalization increases the
demand for productive spending and for lower tax squeezing social
welfare spending. All variables are included in logs. We also introduce
country dummies capturing institutional factors and country preferences
affecting the composition of government spending and a time trend
reflecting different tendencies of each of the functions. The per capita
income (in Purchasing Power Parities of the 2000 dollar and in real
terms of that year), total population, and openness (measured as the sum
of exports and imports divided by the GDP) are obtained from the OECD:
National Accounts: Volume I. Main Aggregates, whereas the age structure
of the population is taken from the OECD: Labour Force Statistics. We
compute permanent income per capita since demand is based on permanent
income rather than on temporary income levels (Peltzman 1980). We
approximate permanent income per capita by taking a three-year moving
average, reducing the sample by two observations for each country.
Introducing dynamics via the lagged dependent variable along with
country dummies is equivalent to demeaning all of the variables by their
individual specific means. This demeaning procedure induces a
correlation between the demeaned lagged dependent variable and the
demeaned error term. (5) Following the Lee and Gordon (2005) analysis on
taxes and growth, we instrument for the lagged share of each function by
the weighted average of the same function in the rest of countries of
the OECD, weighted by the inverse of the distance between capitals
(source: Centre d'Etudes Prospectives et d'Informations
Internationales). This instrument is exogenous, as a country composition
of government spending will not determine the average of the rest of
countries, but it will be influenced by the fiscal policy in the
neighbor countries. Sanz (2006) shows that there is a growing fiscal
interdependence among OECD countries over the period from 1970 to 1997,
particularly for the case of productive spending, which might be
indicating a competition to attract FDI. Borck, Caliendo, and
Steiner (2007) also show significant interaction in almost all spending
subcategories using a cross-section of 435 German counties from 2002,
which they also attribute to fiscal competition among local governments
for mobile factors. Hauptmeier, Mittermaier, and Rincke (2009) find that
if a neighbor provides more infrastructure, governments react by
increasing their own spending on public inputs, for a sample of 1100
German municipalities in the state of Baden-Wurttemberg over the
1998-2004 period. Income per capita also introduces simultaneity, as
government expenditure and its composition affect long-run economic
growth. (6) Furthermore, there might also be a correlation between the
errors and aggregate government spending. Therefore, we instrument
income per capita and the size of aggregate government expenditure by at
least two lagged values of the weighted (by the inverse of the distance)
average income per capita and public sector size of the rest of the OECD
countries.
4. Econometric Results
Table 2 shows the effects of public sector size changes on the
composition of government spending. Appendix 1 provides some diagnostic
testing of our IV regressions. For each case, the Sargan test results do
not reject the hypothesis that the instruments are valid because they
are orthogonal to the error process. However, the instruments should
also be correlated with the included endogenous variables. The usual
F-statistic and the partial [R.sup.2] between all excluded instruments
and the endogenous regressors of the first stage cannot reveal the
weakness of a particular instrument if the rest of the instruments are
highly correlated with the endogenous variables (Staiger and Stock
1997). The Shea partial [R.sup.2] (Shea 1997) overcomes this by taking
into account cross-correlations among the instruments. As a rule of
thumb, Baum, Schaffer, and Stillman (2003) suggest that if the standard
[R.sup.2] is large (whereas the Shea partial [R.sup.2] is small) we may
conclude that the instruments lack sufficient relevance to explain all
the endogenous regressors. Appendix 1 shows that the Shea partial
[R.sup.2] values are all satisfactorily high relative to the usual
[R.sup.2] in each IV regression. A more formal test, which we also
report, is the Stock and Yogo (2005) weak instrument test based on the
Cragg-Donald statistic. This tests whether the bias in IV parameter
estimates due to weak instruments exceeds (above a certain threshold)
the bias in equivalent ordinary least-squares estimates.
For all regressions we can reject the null hypothesis that our
instruments are weak. We show only the average short-run coefficients
across the 25 OECD countries for the sake of space.
Focusing on the variable of interest, we find a negative and
significant relationship in the long run between aggregate government
expenditure and education and transport and communication. Therefore,
reductions (increases) in aggregate government expenditure increase
(reduce) the share devoted to these functions. The result for education
is in line with the results of Dunne, Pashardes, and Smith (1984) for
general current government expenditure in the United Kingdom during the
1950-1980 period and those of Borge and Rattso (1995) for local
expenditure in Norway during the 1986-1989 period, who find that
education expenditures have the property of necessities. Furthermore,
our results are consistent with those of Snyder and Yackovlev (2000),
who find that primary and secondary education spending are relatively
more insulated from economic shocks, and also with those of Cashin et
al. (2001) and Baqir (2002), who find that at a time of budgetary cuts
education is protected. We also find that transport and communication is
isolated from fiscal contractions. On the other hand, social security,
economic affairs, defense, housing, and cultural affairs show a positive
and significant relationship with aggregate government expenditures,
indicating more than proportional reactions to changes in the size of
aggregate government expenditure. Finally, we find that health, public
services, and environment protection react proportionally to changes in
the size of government expenditure. (7)
These results do not entirely support or reject any of the two
hypotheses considered, although they are closer to the second strand of
studies, predicting that budgetary cuts will protect productive
expenditures and fall on social expenditures. We find that transport and
communications, the most productive spending (Easterly and Rebelo 1993),
are protected from budgetary cuts, whereas social security is
particularly targeted. Nevertheless, we find that productive
spending--defense and housing--reduces its share in government spending
at times of fiscal contraction. Finally, we confirm the predictions of
the two hypotheses discussed in section 2 with respect to education.
Since this function is both productive and social, education is isolated
from fiscal contractions.
Most of the rest of the results are in line with the economic
literature. From the speed of adjustment, it can be seen that defense
and education expenditure seem to be the most rigid of the components of
government expenditure. These results are in line with those of Dunne,
Pashardes, and Smith (1984), who find that military expenditure is the
component showing the character of being the most long-planned
expenditure among general current government expenditures in the United
Kingdom over the 1950-1980 period. Income increases the share of
functions such as education and health in the long-run, confirming that
Wagner's law is especially applicable for these two social services
(Peacock and Scott 2000). We also confirm the result found by Shelton
(2007), in which per capita income has no long-run effects on social
security spending, once the elderly share has been controlled for. As
expected, income has a negative association with social security
spending in the short run, confirming that this function is
anti-cyclical. Population reduces the allocation to pure public goods
such as defense (Murdoch and Sandier [1985], for Australian military
demand over the 1961-1979 period), public services, economic services,
and cultural affairs. It also shows a negative association with another
pure public good, such as transport and communications (Randolph,
Bogetic, and Hefley [1996], for 27 low- and middle-income countries over
the 1980-1996 period), although not to a significant extent. The elderly
population increases the demand for health and social security, as this
age group receives more benefits from it than do the other age groups.
This result is partially consistent with that of Lindert (1996), who
finds that an ageing population is strongly and positively associated
with social welfare spending for a panel of 19 OECD countries during the
1960-1981 period. Nevertheless, this author does not find any
significant effect of the elderly population on health expenditure. We
also find evidence of the elderly increasing public services and defense
spending, showing a preference for security-related functions of
government spending. As for the young population, we find a positive and
significant impact on education and health spending. Finally, we find
that openness to international trade increases social spending:
education, health, social security--although the latter not to a
significant extent--along with transport and communications. This result
is partially consistent with that of Shelton (2007), who finds that
openness increases most of the components of government spending, but
particularly education, public order, and safety and transport and
communication, in a sample of 100 countries over the 1970-2000 period.
Nevertheless, Shelton (2007) does not find any particular impact of
trade openness in social security spending, in contrast to the
compensation hypothesis put forward by Rodrik (1998) and tested in
Gemmell, Kneller, and Sanz (2008). Spending on education, economic
affairs, transport and communications, housing, and cultural affairs
shows a negative time trend, whereas health, public services, defense,
and the new environment protection function increase over time. The sign
of the time trend matches the evolution of these functions, except in
the case of defence, a function that has decreased its share in
aggregate government spending since the 1970s. Population, trade
openness, and, above all, increasing per capita income are the variables
driving down defense spending (rather than a time trend).
In Table 2 the effects of fiscal expansions and contractions on the
composition of government expenditure were constrained to be the same.
Nevertheless, components may react differently to increases or decreases
in government spending. In fact, Ravallion (2002) finds that social
spending in Argentina during the 1980s and 1990s decreased more than
proportionally after reductions in government expenditure, whereas it
did not significantly increase following rises in government spending.
Thus, we include a variable that takes the value of the government
spending if the public sector is reduced and that equals zero otherwise.
Table 3 shows results of the estimation, taking into account possible
asymmetry on the effects of government size changes. We choose not to
report the results for the control variables for the sake of space.
These are available from the author on request. Again, the Sargan test
statistic of over-identifying restrictions does not reject the validity
of the instruments used (see Appendix 2). For all regressions we can
reject the null hypothesis that our instruments are weak.
Looking at the coefficients associated with government size, it may
be seen that social security shows a positive and significant
association with the public sector size at fiscal expansion. When
aggregate government spending is decreasing, social security also
decreases. Nevertheless, at times of budgetary cuts, this positive
association is significantly lower than at times of public sector
growth. As indicated by Figure 1, social security spending increases by
a greater amount in fiscal expansions than the extent to which it is
diminished in fiscal contractions. The other two social functions,
education and health, which also represent productive spending, show a
negative relation with the aggregate government spending, particularly
when the public sector size is diminishing. That is, those two functions
are more isolated from fiscal contractions than from fiscal expansions.
Transport and communication is also protected from budgetary cuts,
showing an inverse and symmetric association with the public sector size
changes. Along with social security, fiscal consolidations fall over
economic affairs, defense, housing, and cultural affairs. Environment
protection shows a negative association with government size, as in the
previous table, but positive and significant associations as government
size decreases. The non-significant coefficients associated with
government size may be due to the fact that it is computed from the
budget constraint and hence with some imprecision. Interestingly, we
find that smaller functions show a positive association with government
size, with the exception of environment protection, which is a rather
new component of government spending and hence is still saved from the
cuts. In the long run, governments adjust those small functions that are
less of a necessity. Defense seems to be the most affected by public
sector size decreases. This result is in line with that of Gupta, Mello,
and Sharan (2001), who find that fiscal consolidations fall primarily
over defense expenditure in a sample of 120 countries over the 1985-1998
period, and in contrast to that of Davoodi et al. (1999) and Jonakin and
Stephens (1999), who find that defense expenditure is more protected
when fiscal discipline is implemented (in a sample of 130 countries for
the 1985-1998 period and five Central America countries over the
1975-1993 period, respectively). We also find that short-run reactions
of the components to budgetary cuts are usually the reverse of the
long-run reactions. Confronted with the necessity of adjusting public
spending, governments initially postpone investments in education and
transport and communication, maybe because it is a quicker solution, but
further budgetary cuts protect these two functions. In contrast, fiscal
adjustments does not affect immediately the most rigid category,
defense, and public services, a function in which spending is to a large
extent pre-committed. Smaller functions, such as cultural and
environment protection, are also isolated from budgetary cuts in the
short run, perhaps because of the fact that it is more difficult to
find, at first, a significant amount of funds to save from an
already-small budget.
To sum up, we find that social spending with a productive
character, such as that associated with education and health, along with
the most productive function, transport and communication, are the
categories most protected from fiscal contractions. The other social
spending, social welfare, reduces its share in aggregate government
spending during times of budgetary cuts, but by less than the increase
of years with growth of the public sector size. Fiscal contractions fall
primarily over smaller and less visible functions, such as economic
affairs, defense, housing, and cultural affairs.
5. Robustness Check
Decreases in government spending are usually the result of a fiscal
adjustment that pursues the reduction of public debt and public deficit.
Zaghini (2001) shows that European Union Member States achieved cutbacks
in the 1990s on public deficit through an expenditure reduction policy.
Moreover, Ardagna (2007) shows that fiscal adjustments achieved mainly
by reducing government spending, instead of increasing government
revenues, are more long lasting and growth enhancing. In this section we
check to determine if the effects of public debt and public deficit
reduction on the composition of government spending are similar to those
found in the previous section. Data is built on OECD: National Accounts.
Volume IV: General Government Accounts.
Appendix 3 shows that we cannot reject the validity of the
instruments used in both robustness checks, and Appendix 4 indicates
that we can reject the null hypothesis that our instruments are weak.
Table 4 shows that the effects of reductions in the public debt ratio to
GDP on the composition of government spending are indeed very similar to
those of aggregate government spending. In the long run, education and
health spending increase their share in government spending when the
public debt ratio to GDP is shrinking. This result is in line with that
of Lora and Oliveira (2007), who find a negative association between the
public debt and education and health spending for a panel of 50
developing countries between 1985 and 2003, concluding that orthodoxy in
fiscal policy avoiding public over-indebtedness is the best way to
protect these two functions. Transport and communications shows a
negative correlation with public debt, but not to a significant extent.
These results are in line with those of Mahdavi (2004), who finds that
external public debt in developing countries adversely affects
productive spending, whereas debt relief raises these types of spending.
Social welfare diminishes its share on aggregate government spending
when the public debt is decreasing, but again by a significantly lower
amount than the increases of this function when the public debt is
rising. Results confirm that fiscal consolidations fall primarily over
economic affairs, defense, housing, and cultural affairs, along with
public services, which is again positive, but now to a significant
extent. Environment protection shows a highly negative association with
public debt, and now to a significant extent, confirming that the fact
that it is a new function isolates this function from fiscal
consolidations. We find again that, in the short run, that fiscal
adjustments protect the smallest functions: housing, cultural affairs,
and environment.
Finally, Table 5 introduces public deficit, instead of public debt,
along with the control variables. A positive association between public
deficit (with a positive sign when expenditures are higher than
revenues) and a function of government spending indicates that this type
of spending reduces its share in the public sector size when there is a
fiscal consolidation process in place and the public deficit is
decreasing. Functions protected at times of fiscal adjustment would be
those with a negative association, indicating that this type of spending
increases its share in aggregate government spending when the public
deficit is decreasing. Results confirm previous findings: education and
health, along with transport and communication and environment, are the
functions most protected during times of fiscal adjustment. Social
security spending reduces its share in budgetary cuts, but again to a
lower extent than the increases of this function in fiscal expansions.
The sums of the coefficients of changes in deficit and decreases of
deficit are significantly positive for defense, housing, and cultural
affairs, indicating that these functions, along with public services and
economic affairs, are cut at times of budgetary cuts.
6. Conclusions
This article explores how fiscal consolidation can affect the
composition of government expenditures by analyzing the relationship
between the size of aggregate government expenditure and each of its
functional components in the OECD over the 1970-2007 period. We find
that at times of fiscal consolidation, expenditures with both a social
and a productive character are the most protected. Education and health
exert a robust negative association with the public sector size,
increasing their share in aggregate government spending when countries
are implementing fiscal adjustment. This is consistent with both
functions reducing their spending less than proportionally to aggregate
government spending, maintaining it, or increasing it altogether.
Moreover, fiscal consolidation does not fall primarily on the most
productive expenditure, transport and communications, but on public
services, economic affairs, defense, housing, and cultural affairs.
Hence, it seems as if governments protect the most productive
expenditures or those that are both productive and social. This result
may be evidence of the government's reaction to the voters'
increasing opinion that reducing
productive expenditures harms long-term economic growth. Along
these lines, Ghate and Zak (2002) elaborate a model in which
voters' support for politicians depends on the transfers they
receive and on output growth. Voters may be willing to accept some
reductions in the transfers they desire if they think that this
stimulates output growth. Therefore, governments strike a balance
between utility and economic-growth-enhancing expenditure by protecting
social spending with a productive character the most. The evidence we
find here does not fully support or reject the two hypotheses presented
in the discussion of previous literature. Some utility-enhancing
(growth-enhancing) functions are protected and others are adjusted at
times of budgetary cuts. Therefore, it is more useful to analyze the
full classification of the composition of government expenditure by
functions. This is an important contribution of this article, since most
of the few analyses on the effects of fiscal consolidation on the
composition of government expenditures have focused on particular
functions or on its economic classification.
Nevertheless, these results are also consistent with governments
reducing pure public goods, such as public services, economic services,
defense, and cultural affairs while protecting merit goods, such as
education and health expenditure, when facing budgetary cuts. As pure
public goods are less visible than merit goods, by reducing the former,
governments may be minimizing the political costs of fiscal
consolidations.
Appendix 1: Instrumental Variable Regression Diagnostics (Table 2) (a)
Social
Security Education
Sargan test [chi square] (8) [chi square] (6)
11.043 [0.20] 7.802 [0.25]
Anderson under-identification [chi square] (9) [chi square] (7)
test 140.4 [0.00] 83.2 [0.00]
Weak identification test: 11.40 9.46
Cragg-Donald statistic (b) (CV: 9.01) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.21 (0.25) 0.23 (0.28)
Per capita income 0.38 (0.47) 0.14 (0.40)
Lagged dependent variable 0.19 (0.21) 0.10 (0.27)
Public
Health Services
Sargan test [chi square] (8) [chi square] (8)
6.056 [0.64] 6.056 [0.64]
Anderson under-identification [chi square] (9) [chi square] (9)
test 175.1 [0.00] 175.1 [0.00]
Weak identification test: 17.14 17.14
Cragg-Donald statistic (b) (CV: 9.01) (CV: 9.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.20 (0.22) 0.20 (0.22)
Per capita income 0.30 (0.44) 0.30 (0.44)
Lagged dependent variable 0.19 (0.26) 0.19 (0.26)
Economic
Affairs Transport
Sargan test [chi square] (13) [chi square] (6)
3.048 [0.99] 6.847 [0.34]
Anderson under-identification [chi square] (14) [chi square] (7)
test 213.2 [0.00] 141.7 [0.00]
Weak identification test: 14.54 16.70
Cragg-Donald statistic (b) (CV: 10.14) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.26 (0.31) 0.22 (0.23)
Per capita income 0.35 (0.50) 0.25 (0.26)
Lagged dependent variable 0.26 (0.33) 0.21 (0.21)
Defense Housing
Sargan test [chi square] (5) [chi square] (9)
6.067 [0.30] 10.958 [0.28]
Anderson under-identification [chi square] (6) [chi square] (10)
test 111.3 [0.00] 115.8 [0.00]
Weak identification test: 16.62 10.05
Cragg-Donald statistic (b) (CV: 6.61) (CV: 9.37)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.20 (0.26) 0.19 (0.20)
Per capita income 0.17 (11.24) 0.39 (0.42)
Lagged dependent variable 0.93 (0.98) 0.13 (0.13)
Cultural
Affairs
Sargan test [chi square] (12)
16.593 [0.17]
Anderson under-identification [chi square] (13)
test 146.2 [0.00]
Weak identification test: 10.31
Cragg-Donald statistic (b) (CV: 10.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.26 (0.29)
Per capita income 0.29 (0.46)
Lagged dependent variable 0.19 (0.28)
(a) Values in brackets represent p values.
(b) CV indicates critical value for the weak instrument test
computed by Stock and Yogo (2005) for up to three endogenous
variables, 10% bias of the UIV estimator relative to OLS and
the number of instruments used in each regression.
Appendix 2: Instrumental Variable Regression Diagnostics (Table 3) (a)
Social
Security Education
Sargan test [chi square] (8) [chi square] (6)
10.078 [0.26] 6.872 [0.34]
Anderson under-identification [chi square] (9) [chi square] (7)
test 129.7 [0.00] 83.1 [0.00]
Weak identification test: 10.46 9.45
Cragg-Donald statisticb (CV: 9.01) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.21 (0.25) 0.23 (0.28)
Per capita income 0.38 (0.47) 0.14 (0.40)
Lagged dependent variable 0.18 (0.21) 0.10 (0.27)
Public
Health Services
Sargan test [chi square] (6) [chi square] (8)
2.545 [0.86] 6.056 [0.64]
Anderson under-identification [chi square] (7) [chi square] (9)
test 176.4 [0.00] 175.1 [0.00]
Weak identification test: 21.14 17.14
Cragg-Donald statisticb (CV: 9.01) (CV: 9.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.20 (0.23) 0.20 (0.22)
Per capita income 0.29 (0.43) 0.30 (0.44)
Lagged dependent variable 0.20 (0.26) 0.19 (0.26)
Economic
Affairs Transport
Sargan test [chi square] (13) [chi square] (6)
2.813 [0.99] 7.757 [0.26]
Anderson under-identification [chi square] (14) [chi square] (7)
test 214.8 [0.00] 142.1 [0.00]
Weak identification test: 14.66 16.73
Cragg-Donald statisticb (CV: 10.14) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.27 (0.31) 0.22 (0.23)
Per capita income 0.34 (0.50) 0.25 (0.26)
Lagged dependent variable 0.26 (0.33) 0.25 (0.26)
Defense Housing
Sargan test [chi square] (4) [chi square] (9)
5.330 [0.26] 10.960 [0.28]
Anderson under-identification [chi square] (5) [chi square] (10)
test 111.3 [0.00] 115.3 [0.00]
Weak identification test: 16.60 9.99
Cragg-Donald statisticb (CV: 6.61) (CV: 9.37)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.20 (0.26) 0.19 (0.19)
Per capita income 0.17 (0.23) 0.38 (0.42)
Lagged dependent variable 0.93 (0.98) 0.13 (0.13)
Cultural
Affairs
Sargan test [chi square] (12)
17.185 [0.14]
Anderson under-identification [chi square] (13)
test 149.2 [0.00]
Weak identification test: 10.53
Cragg-Donald statisticb (CV: 10.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government size 0.26 (0.29)
Per capita income 0.30 (0.46)
Lagged dependent variable 0.19 (0.28)
CV indicates critical value.
(a) Values in brackets represent p values.
(b) For economic affairs, the weak instrument test is rejected
for a 20% of maximal bias of the IV estimator relative to OLS.
Instruments rejecting this test for a lower relative bias do not
pass the Sargan test.
Appendix 3: Instrumental Variable Regression Diagnostics (Table 4) (a)
Social
Security Education
Sargan test [chi square] (6) [chi square] (6)
9.623 [0.14] 8.964 [0.18]
Anderson under-identification [chi square] (7) [chi square] (7)
test 83.9 [0.00] 137.7 [0.00]
Weak identification test: 9.57 16.26
Cragg-Donald statistic (CV: 9.01) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government debt 0.15 (0.25) 0.20 (0.25)
Per capita income 0.31 (0.44) 0.29 (0.51)
Lagged dependent variable 0.16 (0.19) 0.25 (0.36)
Public
Health Services
Sargan test [chi square] (10) [chi square] (8)
14.89 [0.14] 5.877 [0.66]
Anderson under-identification [chi square] (11) [chi square] (9)
test 148.3 [0.00] 139.2 [0.00]
Weak identification test: 12.23 12.43
Cragg-Donald statistic (CV: 9.01) (CV: 9.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government debt 0.16 (0.26) 0.18 (0.22)
Per capita income 0.48 (0.67) 0.37 (0.46)
Lagged dependent variable 0.27 (0.37) 0.23 (0.24)
Economic
Affairs Transport
Sargan test [chi square] (8) [chi square] (7)
1.914 [0.98] 6.507 [0.48]
Anderson under-identification [chi square] (9) [chi square] (8)
test 92.6 [0.00] 125.8 [0.00]
Weak identification test: 8.66 13.23
Cragg-Donald statistic (CV: 10.14) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government debt 0.14 (0.25) 0.15 (0.17)
Per capita income 0.22 (0.45) 0.24 (0.27)
Lagged dependent variable 0.15 (0.28) 0.20 (0.21)
Defense Housing
Sargan test [chi square] (5) [chi square] (8)
7.941 [0.16] 7.796 [0.45]
Anderson under-identification [chi square] (6) [chi square] (9)
test 94.3 [0.00] 102.9 [0.00]
Weak identification test: 12.20 9.68
Cragg-Donald statistic (CV: 6.61) (CV: 9.37)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government debt 0.10 (0.15) 0.19 (0.19)
Per capita income 0.23 (0.24) 0.29 (0.44)
Lagged dependent variable 0.63 (0.97) 0.17 (0.24)
Cultural
Affairs
Sargan test [chi square] (11)
9.920 [0.54]
Anderson under-identification [chi square] (12)
test 124.2 [0.00]
Weak identification test: 9.33
Cragg-Donald statistic (CV: 10.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government debt 0.19 (0.25)
Per capita income 0.38 (0.59)
Lagged dependent variable 0.19 (0.28)
CV indicates critical value.
Values in brackets represent p values.
Appendix 4: Instrumental Variable Regression Diagnostics (Table 5) (a)
Social
Security Education
Sargan test [chi square] (6) [chi square] (8)
6.861 [0.33] 12.488 [0.13]
Anderson under-identification [chi square] (7) [chi square] (9)
test 89.9 [0.00] 125.1 [0.00]
Weak identification test: 10.29 11.98
Cragg-Donald statistic (CV: 9.01) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government deficit 0.18 (0.26) 0.23 (0.23)
Per capita income 0.44 (0.44) 0.28 (0.50)
Lagged dependent variable 0.11 (0.16) 0.14 (0.24)
Public
Health Services
Sargan test [chi square] (5) [chi square] (8)
7.322 [0.20] 4.429 [0.82]
Anderson under-identification [chi square] (6) [chi square] (9)
test 109.6 [0.00] 147.4 [0.00]
Weak identification test: 14.33 14.21
Cragg-Donald statistic (CV: 9.01) (CV: 9.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government deficit 0.14 (0.16) 0.17 (0.18)
Per capita income 0.43 (0.57) 0.43 (0.46)
Lagged dependent variable 0.20 (0.30) 0.24 (0.24)
Economic
Affairs Transport
Sargan test [chi square] (8) [chi square] (7)
2.508 [0.96] 5.899 [0.55]
Anderson under-identification [chi square] (9) [chi square] (8)
test 130.8 [0.00] 180.9 [0.00]
Weak identification test: 12.50 19.64
Cragg-Donald statistic (CV: 10.14) (CV: 7.77)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government deficit 0.15 (0.20) 0.15 (0.17)
Per capita income 0.36 (0.46) 0.26 (0.26)
Lagged dependent variable 0.17 (0.27) 0.22 (0.22)
Defense Housing
Sargan test [chi square] (5) [chi square] (8)
4.663 [0.46] 7.890 [0.44]
Anderson under-identification [chi square] (6) [chi square] (9)
test 83.6 [0.00] 159.2 [0.00]
Weak identification test: 10.75 15.47
Cragg-Donald statistic (CV: 6.61) (CV: 9.37)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government deficit 0.09 (0.16) 0.19 (0.20)
Per capita income 0.16 (0.24) 0.39 (0.44)
Lagged dependent variable 0.61 (0.97) 0.16 (0.17)
Cultural
Affairs
Sargan test [chi square] (11)
8.860 [0.64]
Anderson under-identification [chi square] (12)
test 196.1 [0.00]
Weak identification test: 15.38
Cragg-Donald statistic (CV: 10.01)
Endogenous variables: Shea
partial [R.sup.2]
(regression [R.sup.2])
Government deficit 0.36 (0.38)
Per capita income 0.46 (0.60)
Lagged dependent variable 0.24 (0.28)
CV indicates critical value.
(a) Values in brackets represent p values.
References
Akitoby, Bernardin, Benedict Clements, Sanjeev Gupta, and Gabriela
Inchauste. 2006. Public spending, voracity, and Wagner's Law in
developing countries. European Journal of Pofitical Economy 22:908-24.
Alesina, Alberto, Roberto Perotti, and Jose Tavares. 1998. The
political economy of fiscal adjustments. Brokings Papers on Economic
Activity Spring 1:197 266.
Ardagna, Silvia. 2007. Determinants and consequences of fiscal
adjustments. Unpublished paper, Harvard University.
Aubin, Christian, Jean P. Berdot, Daniel Goyeau, and Jean D. Lafay.
1988. The growth of public expenditure in France. In Explaining the
growth of government, edited by Johan A. Lybeck and Magnus Henrekson.
Amsterdam: Elsevier Science Publishers, pp. 201-30.
Baqir, Reza. 2002. Social sector spending in a panel of countries.
International Monetary Fund Working Paper WP/02/35.
Baum, Christopher, Mark E. Schaffer, and Steven Stillman. 2003.
Instrumental variables and GMM: Estimation and testing. Boston College
Working Paper No. 545.
Beck, Nathaniel, and Jonathan N. Katz. 2009. Modeling dynamics in
time-series-cross-section political economy data. California Institute
of Technology Social Science Working Paper No. 1304.
Borcherding, Thomas E., Stephen F. Ferris, and Andrea Garzoni.
2004. Growth in the real size of government since 1970. In Handbook of
public finance, edited by Richard E. Wagner and Jtirgen G. Backhaus.
Dordrecht: Kluwer Academic Publisher, pp. 77-108.
Borck, Rainald, Marco Caliendo, and Viktor Steiner. 2007. Fiscal
competition and the composition of public spending: Theory and evidence.
Finanzarchiv/Public Finance Analysis 63:264-77.
Borge, Lars E., and Jorn Rattso. 1995. Demographic shift, relative
costs and the allocation of local government consumption in Norway.
Regional Science and Urban Economics 25:705-26.
Cashin, Paul A., Paolo Mauro, Catherine A. Pattillo, and Ratna
Sahay. 2001. Macroeconomic policies and poverty reduction: Stylized
facts and an overview of research. International Monetary Fund Working
Paper 01/135.
Davoodi, Hamid, Benedict Clements, Jerald Schiff, and Peter
Debaere. 1999. Military spending, the peace dividend, and fiscal
adjustment. International Monetary Fund Working Paper WP/99/87.
Davoodi, Hamid, and Heng-fu Zou. 1998. Fiscal decentralization and
economic growth: A cross country study. Journal of Urban Economics
43:244-57.
de Mello, Luiz. 2008. Estimating a fiscal reaction function: The
case of debt sustainability in Brazil. Applied Economics 40:271-84.
Devarajan, Shantayanan, Vinaya Swaroop, and Heng-fu Zou. 1996. The
Composition of public expenditure and economic growth. Journal of
Monetary Economics 37:31344.
Drazen, Allan, and Marcela Eslava. 2010. Electoral manipulation via
expenditure composition: Theory and evidence. Journal of Development
Economics 92:39-52.
Dunne, John P., Panos Pashardes, and Ronald P. Smith. 1984. Needs,
costs and bureaucracy: The allocation of public consumption in the UK.
Economic Journal 94:1-15.
Easterly, William, Timothy Irwin, and Luis Serv6n. 2008. Walking up
the down escalator: Public investment and fiscal stability. Worm Bank
Research Observer 23:37-56.
Easterly, William, and Sergio Rebelo. 1993. Fiscal policy and
economic growth. Journal of Monetary Economics 32:417 58.
Gemmell, Norman, Richard Kneller, and Ismael Sanz. 2008. Foreign
investment, international trade and the size and structure of public
expenditures. European Journal of Pofitical Economy 24:151-71.
Gemmell, Norman, Richard Kneller, and Ismael Sanz. 2009. The
composition of government expenditure and economic growth: Some evidence
from OECD countries. European Economy Occasional Papers 45.
Ghate, Chetan, and Paul J. Zak. 2002. Growth of government and the
politics of fiscal policy. Structural Change and Economic Dynamics
13:435-55.
Gomes, Pedro, and Francois Pouget. 2008. Corporate tax competition
and the decline of public investment. European Central Bank Working
Paper Series No. 928.
Gupta, Sanjeev, Luiz Mello, and Raju Sharan. 2001. Corruption and
military spending. European Journal of Political Economy 17:749 77.
Hauptmeier, Sebastian, Ferdinand Mittermaier, and Johannes Rincke.
2009. Fiscal competition over taxes and public inputs. Theory and
evidence. European Central Bank Working Paper Series No. 1033.
Henrekson, Magnus. 1988. Swedish government growth. A
disequilibrium analysis. In Explaining the growth of government, edited
by Johan A. Lybeck and Magnus Henrekson. Amsterdam: Elsevier Science
Publishers, pp. 93-132.
International Monetary Fund (IMF). 2002. Government finance
statistics manual 2001. Companion material. Washington: IMF.
Jonakin, Jon, and Mark Stephens. 1999. The impact of adjustment and
stabilization policies on infrastructure spending in Central America.
North American Journal of Economics and Finance 10:293 308.
Keen, Michael, and Maurice Marchand. 1997. Fiscal competition and
the pattern of public spending. Journal of Public Economics 66:33 53.
Kneller, Richard, Michael Bleaney, and Norman Gemmell. 1999. Fiscal
policy and growth: Evidence from OECD countries. Journal of Public
Economics 74:171-90.
Lee, Young, and Roger H. Gordon. 2005. Tax structure and economic
growth. Journal of Public Economics 89:1027-43.
Lindert, Peter H. 1996. What limits social spending? Explorations
in Economic History 33:1-34.
Lora, Eduardo, and Mauricio Olivera. 2007. Public debt and social
expenditure: Friends or foes? Emerging Markets Review 8:299-310.
Mahdavi, Saeid. 2004. Shifts in the composition of government
spending in response to external debt burden. Worm Development 32: l
139-57.
Murdoch, James C., and Todd Sandier. 1985. Australian demand for
military expenditures: 1961-1979. Australian Economic Papers 24:142-53.
Oxley, Howard, and John P. Martin. 1991. Controlling government
spending and deficit: Trends in the 1980s and prospects for the 1990s.
OECD Economic Studies 17:145-89.
Peacock, Alan T., and Alex Scott. 2000. The curious attraction of
Wagner's law. Public Choice 102:1 17.
Peltzman, Sam. 1980. The growth of government. Journal of Law and
Economics 23:209-88.
Pesaran, M. Hashem, Yongcheol Shin, and Ron Smith. 1999. Pooled
mean group estimation of dynamic heterogeneous panels. Journal of the
American Statistical Association 94:621-34.
Randolph, Susan, Zeljko Bogetic, and Dennis Hefley. 1996.
Determinants of public expenditure on infrastructure. Transportation and
communication. World Bank Policy Research Working Paper 1661.
Ravallion, Martin. 1999. Is more targeting consistent with less
spending? International Tax and Public Finance 6:411-9.
Ravallion, Martin. 2000. Are the poor protected from budget cuts?
Theory and evidence for Argentina. World Bank Policy Research Working
Paper 2391.
Ravallion, Martin. 2002. Are the poor protected from budget cuts?
Evidence for Argentina. Journal of Applied Economics 5:95 121.
Rodrik, Dani. 1998. Why do more open economies have bigger
governments? Journal of Political Economy 106:997-1032.
Roubini, Nouriel, and Jeffrey Sachs. 1989. Government spending and
budget deficits in the industrial countries. Economic Policy 8:99-132.
Sanz, Ismael. 2006. Is there any race to the top of productive
government expenditures? Evidence from OECD countries (1970-1997).
European Political Economy Review 3:29-52.
Sanz, Ismael, and Francisco J. Velazquez. 2007. The role of ageing
in the growth of government and social welfare spending in the OECD.
European Journal of Political Economy 23:917-31.
Shea, John. 1997. Instrument relevance in multivariate linear
models: A simple measure. Review of Economics and Statistics 79:348-52.
Shelton, Cameron A. 2007. The size and composition of government
expenditure. Journal of Public Economics 91:2230-60.
Snyder, James M., and Irene Yackovlev. 2000. Political and economic
determinants of government spending on social protection programs.
Unpublished paper, Massachusetts Institute of Technology.
Staiger, Douglas, and James H. Stock. 1997. Instrumental variables
regression with weak instruments. Econometrica 65:557-86.
Stock, James H., and Motohiro Yogo. 2005. Testing for weak
instruments in linear IV regression. In Identification and inference for
econometric models: Essays in honor of Thomas Rothenberg, edited by D.
W. K. Andrews and James H. Stock. Cambridge, UK: Cambridge University
Press, pp. 80-108.
Sturm, Jan E. 1998. Public capital expenditure in the OECD
countries: The causes and impact of the decline in public capital
spending. Cheltenham, UK: Edward Elgar Publishing.
Tanzi, Vito. 2000. Globalisation and the future of social
protection. International Monetary Fund Working Paper No. 12.
Tanzi, Vito, and Ludger Schuknecht. 1997. Reforming government: An
overview of recent experience. European Journal of Political Economy
13:395-417.
Tanzi, Vito, and Ludger Schuknecht. 2000. Public spending in the
20th century." A global perspective. Cambridge, UK: Cambridge
University Press.
Tridimas, George. 2001. The economics and politics of the structure
of public expenditure. Public Choice 106:299-316.
United Nations. 1981. Classification of the functions of
government. New York: United Nations.
United Nations. 2000. Classification of the functions of government
(COFOG). New York: United Nations.
Zaghini, Andrea. 2001. Fiscal adjustments and economic performing:
A comparative study. Applied Economics 33:613-24.
(1) For extension of the new category of environment protection we
have used the variation rates of the former housing and community
amenity. We have also detracted the new public debt transactions from
general public services and extend it back using the variation rates of
other non-classified functions. In this way we can distinguish interest
payments from the rest of the general public services. We have further
summed up general public services (excluding interest payments) and
public order and safety under the category of public services, since
some countries provide information on these functions aggregated and
because the two categories correspond to a similar concept of pure
public goods.
(2) We use the OECD National Accounts inasmuch as it offers
information on the consolidated spending of all levels of government;
additionally, it follows the accrual criterion. Accrual accounting
systems record transactions at the time the economic value is created
rather than when cash transactions take place. This provides a better
picture of the commitments undertaken by governments than does
traditional cash accounting. IMF, Government Finance Statistics, also
provide information related to government spending using the COFOG
classification. This covers a longer time period but is generally
focused on central government spending and is measured on a cash basis.
The OECD data set contains some missing years, and we use the IMF data
to fill in the missing observations.
(3) Social protection is relevant for the utility function but is
considered as non-productive, as it is not privately productive and does
not enter the production function. In fact, Gemmell, Kneller, and Sanz
(2009) do not find any significant effect (positive or negative) of
social protection spending on economic growth. Housing, environment
protection, and recreation, culture, and religion are considered as
non-social functions since these spending types are mostly devoted to
intermediate consumption and gross fixed capital formation rather than
social transfers or social benefits in kind (see Eurostat, Statistics in
Focus: 28/05).
(4) We have to assume that relative public prices between functions
remain constant over the 1970-2007 period. There are no data available
on prices for functions of government expenditure. The difficulty of
obtaining the prices of the public goods of each type of function lies
in the fact that most of the previous empirical studies analyzing
particular functions do not include prices (Shelton 2007).
(5) This bias gets smaller as we increase the number of years. We
control for endogeneity, though Beck and Katz (2009) show that the Least
Squares Dummy Variables lead to relatively good performance when T is 20
or more, in terms of bias and root mean square error.
(6) See Gemmell, Kneller, and Sanz (2009) for a survey on fiscal
policy and economic growth. This endogenous issue has not received much
attention in empirical studies (Borcberding, Ferris, and Garzoni 2004).
This may be because this potential source of endogeneity is reduced,
considering that the size and composition of public spending effects on
economic growth are likely to unfold slowly (Devarajan, Swaroop, and Zou
1996).
(7) Coefficients for environment protection are computed from the
budget constraint. Because of this computation and because environment
protection is a small function with great percentage changes, some of
the coefficients estimated are very high.
Table 1. Classification of Government Expenditures by Function
Former COFOG New COFOG Kneller et
UN (1981) UN (2000) al. (1999) Ravallion (2002)
Economic affairs Economic affairs
Recreation, Recreation, Non-productive Non-social
culture and culture and
religion religion
Social security Social security Social
and welfare and welfare
Education Education
Health Health
Housing and Housing and Productive
community community Non-social
amenities amenities
Defense Environment
Transport and protection
communication Defense
Public order and Transport and
safety communication
General public Public order
services and safety
Other non- General public
classified services
functions General public Other non- Other non-
services classified classified
(public debt functions functions
transactions)
Table 2. Government Size and the Composition of
Government Expenditure by Function (1970-2007)
Social Public
Security Education Health Services
Level effects
Government 0.098 -0.466 -0.075 0.002
size (1.77) * (6.68) *** (1.14) (0.04)
Per capita -0.040 0.089 0.383 0.037
income (1.21) (1.85) * (4.45) *** (0.70)
Total 0.237 -0.013 -0.876 -0.341
population (1.64) (0.15) (8.92) *** (2.19) **
Share of 0.138 0.033 0.641 0.581
elderly in (1.97) ** (0.43) (4.63) *** (6.60) ***
population
Share of -0.181 0.414 0.390 0.253
young in (2.69) *** (6.58) *** (6.69) *** (4.33) ***
population
Openness to 0.053 0.231 0.104 0.000
international (1.10) (4.73) *** (2.83) *** (0.00)
trade
Trend -0.001 -0.006 0.008 0.005
(0.39) (3.28) *** (4.84) *** (3.20) ***
Adjustment -0.403 -0.344 -0.414 -0.419
parameter (4.80) *** (6.48) *** (4.01) *** (7.34) ***
Short-run effects
Government -0.105 0.141 -0.315 0.018
size (1.50) (0.85) (2.62) *** (0.06)
Per capita -0.141 -0.018 -0.195 0.070
income (3.32) *** (0.23) (0.89) (0.67)
Total -0.743 -4.580 -1.415 10.922
population (0.51) (0.93) (0.74) (1.03)
Share of 1.381 1.829 1.346 -3.099
elderly in (0.79) (1.05) (0.97) (2.49) **
population
Share of 3.127 0.484 -0.109 -0.701
young in (0.95) (1.03) (0.10) (0.29)
population
Openness to -0.085 -0.031 0.019 -0.221
international (1.23) (0.51) (0.43) (2.52) **
trade
Constant -0.093 0.453 3.929 1.870
(1.62) (5.60) *** (4.02) *** (7.37) ***
Economic
Affairs Transport Defense Housing
Level effects
Government 0.239 -0.323 0.501 0.418
size (2.30) ** (2.89) *** (2.44) ** (1.95) *
Per capita 0.226 -0.227 -3.112 0.309
income (3.22) *** (2.23) ** (9.47) *** (2.37) **
Total -0.360 -0.160 -0.494 4.464
population (2.46) ** (0.74) (1.78) * (13.40) ***
Share of 0.140 0.247 0.576 -0.562
elderly in (1.35) (1.55) (2.21) ** (2.74) ***
population
Share of -0.431 -0.504 -1.493 -1.581
young in (3.73) *** (3.97) *** (5.95) *** (8.58) ***
population
Openness to -0.245 0.327 -1.093 0.069
international (4.42) *** (5.97) *** (9.39) *** (0.92)
trade
Trend -0.011 -0.022 0.041 -0.069
(4.66) *** (8.20) *** (7.06) *** (12.73) ***
Adjustment -0.506 -0.519 -0.029 -0.427
parameter (8.11) *** (8.72) *** (22.28) *** (5.83) ***
Short-run effects
Government 0.174 0.014 -0.040 0.197
size (0.92) (0.09) (3.08) *** (1.07)
Per capita 0.212 -0.360 0.146 0.034
income (1.58) (1.71) * (11.69) *** (0.20)
Total -9.516 7.976 -1.041 -4.438
population (1.30) (1.25) (1.40) (0.79)
Share of 5.831 -2.173 0.030 2.382
elderly in (1.00) (1.44) (0.19) (0.53)
population
Share of 0.381 -2.260 -0.083 6.856
young in (0.19) (1.09) (0.60) (1.17)
population
Openness to 0.029 0.015 0.030 0.343
international (0.20) (0.10) (5.58) *** (1.67) *
trade
Constant 3.343 4.215 1.390 -29.961
(7.31) *** (8.56) *** (21.84) *** (5.74) ***
Cultural
Affairs Environment
Level effects
Government 0.635 -0.738
size (10.66) *** (1.07)
Per capita 1.642 3.492
income (15.17) *** (4.63) ***
Total -0.801 2.340
population (3.27) *** (1.89) *
Share of 1.550 -16.976
elderly in (8.54) *** (18.54) ***
population
Share of 0.980 5.626
young in (5.72) *** (7.25) ***
population
Openness to -0.258 -0.211
international (3.61) *** (0.52)
trade
Trend -0.014 0.119
(3.56) *** (5.69) ***
Adjustment -0.428 --
parameter (6.01) ***
Short-run effects
Government 0.034 2.260
size (0.22) (1.90) **
Per capita 0.375 3.495
income (1.30) (3.54) ***
Total -6.619 25.324
population (0.64) (0.62)
Share of 3.810 -63.230
elderly in (0.78) (3.33) ***
population
Share of 12.154 -84.447
young in (0.85) (3.29) ***
population
Openness to 0.141 2.320
international (0.78) (3.26) ***
trade
Constant -5.028 --
(6.40) ***
t statistics in parentheses.
* Significant at 10%.
** Significant at 5%.
*** Significant at 1%.
Table 3. Asymmetric Effects of Government Size Changes
in the Composition of Government Spending by Function (a)
Social Public
Security Education Health Services
Level effects
Government 0.127 -0.509 -0.115 0.040
size (1.81) * (7.27) *** (1.26) (0.59)
Government -0.012 -0.005 -0.010 0.002
size decreases (5.78) *** (2.47) ** (3.52) *** (0.96)
Short-run effects
Government 0.016 0.005 -0.366 0.092
size (1.76) * (0.03) (1.50) (0.32)
Government 0.001 0.002 0.001 -0.004
size decreases (9.90) *** (1.84) * (1.23) (1.91) *
Economic
Affairs Transport Defense Housing
Level effects
Government 0.222 -0.603 0.503 0.512
size (2.05) ** (4.65) *** (2.50) ** (2.26) **
Government -0.004 -0.002 0.018 -0.017
size decreases (1.40) (0.61) (3.69) *** (2.94) ***
Short-run effects
Government 0.159 0.121 -0.014 -0.111
size (0.84) (0.91) (1.07) (0.48)
Government 0.001 0.003 -0.001 0.002
size decreases (0.44) (1.86) * (5.91) *** (0.81)
Cultural
Affairs Environment
Level effects
Government 1.035 -0.454
size (10.11) *** (0.59)
Government 0.007 0.346
size decreases (1.43) ** (16.09) ***
Short-run effects
Government -0.059 0.938
size (0.35) (0.73)
Government -0.004 -0.029
size decreases (-2.18) ** (2.99) ***
t statistics in parentheses.
(a) Also included in the regression are country fixed effects and
time trend, the measures of per capita income, the total
population, the share of the elderly and young in the total"
population, and openness to international trade, all in levels and
first differences.
* Significant at 10%.
** Significant at 5%.
*** Significant at 1%.
Table 4. Asymmetric Effects of Public Debt Changes in
the Composition of Government Spending by Function (a)
Social Public
Security Education Health Services
Level effects
Public debt 0.055 -0.018 -0.080 0.071
ratio (3.13) ** (0.75) (2.71) ** (2.38) *
Decreases in -0.003 -0.004 -0.004 -0.002
public debt (2.04) * (2.61) ** (2.01) * (0.91)
Short-run effects
Public debt 0.029 -0.044 0.004 -0.093
ratio (1.00) (1.85) (0.16) (0.95)
Decreases in -0.001 0.003 0.002 0.001
public debt (0.53) (2.73) ** (1.50) (0.21)
Economic
Affairs Transport Defense Housing
Level effects
Public debt 0.082 -0.038 0.204 0.127
ratio (2.27) * (0.85) (2.65) ** (2.06) *
Decreases in 0.005 -0.001 -0.005 -0.003
public debt (1.75) (0.25) (1.34) (0.44)
Short-run effects
Public debt 0.062 0.020 -0.080 -0.132
ratio (1.19) (1.19) (1.58) (2.41) *
Decreases in -0.001 0.001 0.001 0.002
public debt (0.91) (0.44) (0.23) (1.15)
Cultural
Affairs Environment
Level effects
Public debt 3.030 -6.241
ratio (7.89) ** (11.96) **
Decreases in 0.065 0.048
public debt (5.33) ** (2.01) *
Short-run effects
Public debt -0.038 0.479
ratio (7.60) ** (1.38)
Decreases in 0.000 -0.031
public debt (1.89) (1.70) *
t statistics in parentheses.
(a) Also included in the regression are country fixed effects and
time trend, the measures of per capita income, the total
population, the share of the elderly and young in the total
population, and openness to international trade, all in levels
and first differences.
* Significant at 10%.
** Significant at 5%.
Table 5. Asymmetric Effects of Public Deficit Changes
in the Composition of Government Spending by Function (a)
Social Public
Security Education Health Services
Level effects
Deficit to 0.030 -0.005 0.004 -0.001
GDP ratio (6.19) *** (2.32) ** (1.29) (0.54)
Decreases in -0.005 -0.001 -0.004 0.004
deficit to (1.71) * (0.31) (2.46) ** (1.65) *
GDP ratio
Short-run effects
Deficit to -0.009 -0.002 -0.001 0.003
GDP ratio (2.24) ** (0.67) (0.45) (0.61)
Decreases in 0.001 0.002 0.001 -0.003
deficit to (0.73) (0.91) (0.77) (1.24)
GDP ratio
Economic
Affairs Transport Defense Housing
Level effects
Deficit to 0.010 -0.022 -0.008 0.048
GDP ratio (1.65) * (4.40) *** (2.77) *** (4.71) ***
Decreases in 0.000 0.000 0.013 -0.018
deficit to (0.05) (0.06) (4.87) *** (3.24) ***
GDP ratio
Short-run effects
Deficit to 0.004 0.011 -0.001 -0.003
GDP ratio (0.73) (2.75) *** (0.46) (0.21)
Decreases in 0.002 0.003 -0.002 -0.001
deficit to (0.93) (1.25) (1.60) (0.15)
GDP ratio
Cultural
Affairs Environment
Level effects
Deficit to 0.022 -0.657
GDP ratio (1.91) * (18.24) ***
Decreases in -0.010 0.120
deficit to (1.72) * (5.68) ***
GDP ratio
Short-run effects
Deficit to 0.072 0.033
GDP ratio (2.73) *** (0.72)
Decreases in 0.001 -0.040
deficit to (1.45) (2.52) **
GDP ratio
t statistics in parentheses.
(a) Also included in the regression are country fixed effects and
time trend, the measures of per capita income, the total
population, the share of the elderly and young in the total
population, and openness to international trade, all in levels
and first differences.
* Significant at 10%.
** Significant at 5%.
*** Significant at 1%.
Ismael Sanz, Applied Economics I, Universidad Rey Juan Carlos,
Campus de Vicalvaro, Paseo artilleros s/n, 28032 Madrid, Spain; E-mail
Ismael.Sanz@urjc.es.
I am grateful to Theodore Bergstrom, Norman Gemmell, and Richard
Kneller for their helpful and interesting comments and suggestions on
this article. Part of this research was undertaken while I was a
Visiting Research Fellow of the Economics Department of the University
of California, Santa Barbara. Financial support from the Science and
Technology Minister of Spain (SEJ2007-66520/ECON) is gratefully
acknowledged.
Received July 2006; accepted January 2010.
