assignment 2
Applied Economics, 2006, 38, 1181–1192
Growth effects of public
expenditure on the state and
local level: evidence from
a sample of rich governments
Christoph A. Schaltegger a, * and Benno Torgler
b
a Swiss Federal Tax Administration, CREMA, Center for Research in
Economics, Management and the Arts, University of St. Gallen, Switzerland b Yale Center for International and Area Studies, Leitner Program
in International & Comparative Political Economy and CREMA,
Center for Research in Economics, Management and the Arts, USA
There is a vast empirical literature investigating the relationship between
government size and economic growth. But the empirical evidence of growth
effects of public expenditure using cross-country regressions is still
inconclusive. According to a number of authors this is not surprising
since the negative relationship only applies for rich countries with a large
public sector. Restricting their analysis on rich countries only they can show
the predicted negative impact. Naturally, a selection of a sub-sample
of rich countries is always somewhat arbitrary. Another possibility is to
concentrate on governments within a rich country. However, only few
studies investigate the effect of state and local spending on economic
growth. This study concentrates on the relationship between public
expenditure and economic growth within a rich country using the full
sample of state and local governments from Switzerland over the 1981–2001
period. The general finding is a fairly robust negative relationship between
government size and economic growth. However, in contrast to public
spending from operating budgets there is no significant impact on economic
growth by expenditure from capital budgets.
I. Introduction
A common feature of all industrialized countries con-
cerns an enormous expansion of the public sector.
As measured by the share of GDP going to govern-
ment expenditures the average OECD country has
expanded its size of government for more than
20 percentage points between 1960 and today.
Today, average government outlays in the OECD
countries account for about nearly 50% of GDP.
Such an enormous government involvement has
attracted various critics including the argument
*Corresponding author. E-mail: Christoph.Schaltegger@estv.admin.ch
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online � 2006 Taylor & Francis 1181 http://www.tandf.co.uk/journals DOI: 10.1080/00036840500392334
of endangering economic prosperity. 1 In particular,
the influential empirical work by Barro (1991) cover- ing a large cross-section of countries supported the view that a large public sector impedes economic growth. Others provided further empirical evidence confirming the negative impact of the size of govern- ment on economic growth (Engen and Skinner, 1992; Hansson and Henrekson, 1994; Grier, 1997; Fölster and Henrekson, 1999, 2001; Romero de Ávila and Strauch, 2003; Bernholz, 2004). However, some authors are very sceptical about the robustness of the provided result. Hsieh and Lai (1994), Atkinson (1995), Slemrod (1995, 1998) or Agell et al. (1997, 1999) find no stable negative correlation between the size of government and economic growth.
Traditionally, theoretical arguments on the impact of fiscal policy on economic growth are split into tax and expenditure policy based aspects (Tanzi and Zee, 1997).
2 Taxation induces distortions in economic
behaviour resulting in allocative inefficiencies as long as lump-sum levies are not feasible. Consequently, the excess burden of taxation adversely affects the long-run growth of an economy. From a dynamic perspective taxation additionally impedes growth by creating disincentives to accumulate physical and human capital depending on its sensitivity in the production technology of the economy. The same applies for the negative impact on labour supply retarding economic growth. Thus, both the level of taxes as well as the structure of the tax system have consequences on economic growth. On the expenditure side, growth effects are more ambiguous. On the one hand, public expenditures are welfare improving if the benefits from these spending exceed private opportunity costs. Furthermore, government activities to secure property rights, to enforce con- tracts and to guarantee a stable monetary regime provide the foundation for a smooth operation
of a market economy. 3 On the other hand, while
public spending can stimulate private sector produc- tivity by the externality of the provided public good, it may also crowd-out private investments and production.
4 Additionally, there is a positive correla-
tion between growth-retarding rent-seeking activities by pressure groups and the size of government expenditures. In a dynamic context, government spending improves private capital accumulation and consequently growth only if being complementary but not substitutive to private expenditures. The opposing driving forces of public expenditures on economic growth point to the importance of the structure of public expenditure. Clearly, while pro- ductive government spending is considered as being growth-inducing, unproductive government activities are expected to be growth-retarding. The empirical challenge here is to reasonably distinguish public expenditure between the two sorts of activities. Thus summing up, the net impact on aggregate output by government expenditures is the sum of both of these effects, growth-inducing and growth-retarding.
The study concentrates on the impact of public expenditures on growth. According to Slemrod (1995), Tanzi and Zee (1997) or Tanzi and Schuknecht (2000), one should only expect a negative impact of the size of government on economic growth if the size of government exceeds a certain threshold. In the US literature, the n-shaped relation between govern- ment size and economic growth is often called the ‘Armey-curve’,
5 according to Richard Armey,
a Member of the House of Representatives (Vedder and Gallaway, 1998). The rationale behind this argument is that in countries with big governments, the share of public expenditures designed to promote private sector productivity is typically smaller than in countries with small governments (Fölster and Henrekson, 2001). For less developed countries,
1 A related question is whether more open economies have a significantly higher government size. Since this study uses state-
level data in a common market, all governments ‘ipso jure’ have the same degree of openness. Hence, the study does not engage in providing evidence on that particular issue. For evidence on the relationship between openness and government size, see Islam (2004). 2 Another fiscal variable that could have an impact on long-run economic growth is the budget balance. First, huge fiscal
imbalances could create political instability resulting in lower economic growth. Second, if the private sector does not react with increased savings, national savings will be reduced by fiscal imbalance and growth hampered. 3 In fact, Keefer and Knack (1997) provide evidence that a legal system protecting property rights and enforcing contracts
enhances economic growth. 4 According to Alesina et al. (2002), the negative effect of public spending is due to an increase in labour costs and thus
a reduction of profits resulting in an investment decline. Their empirical analysis reports quite sizeable effects for OECD countries over the 1960–1996 period. 5 The graphical illustration is borrowed by Arthur Laffer’s ‘Laffer-curve’, indicating the same relationship between tax rates
and tax revenues. The n-shaped relationship has a crucial impact for the empirical analysis since it stresses the importance of an appropriate sample selection.
1182 C. A. Schaltegger and B. Torgler
government spending may act as a signal that
property rights will be enforced. 6 In this case, an
increase of the size of government is likely not to
hamper economic growth. Thus, small government
by itself is not an asset. When a small government
fails to protect property rights and to enforce
contracts, there is no reason to believe that it will
promote economic growth (Gwartney et al., 1998).
However, it is a narrow path to the point where
a growing size of government reflects excessive
engagements in transfer programmes and regulations
that are growth impeding. As stated by Weingast
(1995, p. 1): ‘The fundamental political dilemma of an
economic system is this: A government strong enough
to protect property rights and enforce contracts
is also strong enough to confiscate the wealth of
its citizens’. There is a vast empirical literature investigating the
relationship between government size and economic
growth for OECD countries. However, according
to Fölster and Henrekson (2001), analysing the
impact of the size of government on economic
growth for a sub-sample of rich countries separately
may give a more detailed picture on the issue due
to the non-monotonic relationship (n-shaped ‘Armey-
curve’). In contrast, results from countries all
with very different degrees of economic development
might be inconclusive since the governments lay on
the upward as well as on the downward sloping
branch of the ‘Armey-curve’. A common approach
to circumvent such a bias is to use a sub-sample of
rich countries. Naturally, a selection of a sub-sample
of rich countries is always somewhat arbitrary. 7
Therefore, an analysis of a full set of sub-federal
governments with considerable autonomy in fiscal
policy decisions within a country represents a possible
alternative. The advantage of such a model selection
is that one can focus on rather homogeneous
governments with respect to their economic develop-
ment, their accounting standards and their legal
framework. But according to the authors’ knowledge
only few authors have been concerned with growth
effects on the sub-federal level. Exceptions are
Helms (1985), Vedder and Gallaway (1998)
Rappaport (1999) or Holcombe and Lacombe
(2004) with evidence from the US state or local level. This study investigates growth effects of govern-
ment spending within a rich country on the state
level. The sample consists of all state governments in
Switzerland, the cantons, over the 1981–2001 period. 8
Analysing growth effects within Switzerland is
reasonable for several reasons. First, the state level
in Switzerland enjoys considerable fiscal autonomy
(Feld et al., 2003). This is especially true for the tax
and expenditure policy. Cantons are free to set tax
rates, tax tariffs, tax exemptions, tax deductions, to
borrow and to spend to a far extent. Second, state
governments in Switzerland have the legal instru-
ments to conduct their own economic policy. The federal government has only very limited pos-
sibility to interfere with policy decisions of cantons.
Compared with other countries’ sub-federal govern-
ments, all state governments within Switzerland can be
considered as rich in terms of their GDP per capita.
Third, accounting standards for governments are
harmonized in Switzerland. This augments the
comparability of our data on public spending of the
cantons. In contrast, different accounting standards
between countries might be very difficult to isolate
in a cross-country analysis. Lastly, all Swiss cantons
separate public spending for current purposes from
spending for investments. Thus, a distinguishing
feature of this study is that one can separate the
effect of public expenditure from operating budgets
to those of capital budgets. The results indicate that the government size
significantly retards economic growth when spending
is used for payments in the operating budgets, while
payments in the capital budget have no significant
effect on economic growth rates. These findings
underscore the importance of different incentives
provided by different spending policies on economic
growth. The paper is organized as follows. Section II
presents some stylized facts on our database and
conduct the empirical analysis while Section III
discusses the obtained results from the regressions.
Finally, Section IV concludes.
6 However, according to de Soto (2002) even though many developing countries face small governments measured by public
spending per GDP they do not necessarily direct the spending in productive government activities. Thus typical problems involved with big governments can be observed in developing countries, too, like over-regulation, interventionism, corruption or bureaucratic slack. 7 Agell et al. (2003, p. 363) argue that the results of cross-country regressions have to be interpreted with caution due to
methodological reasons: ‘A policy-maker who wants to promote growth is well-advised to look for other evidence than cross-country regressions’. 8 There exists a time series analysis for Switzerland by Singh and Weber (1997) concluding that there is no clear empirical
evidence on growth effects by government spending.
Growth effects of public expenditure 1183
II. The Size of Swiss State and Local Governments and Economic Growth
In the past years a number of Swiss cantons have implemented budget rules in response to the revenue shortfall of the early 1990s (Schaltegger, 2002). With one exception (canton Ticino), the rules are designed as balanced-budget-requirements but differ- ing in their strictness to enforce the requirements. Although much of the public debate since then is circling around preferences of tax and expenditure combinations, the questions of how state and local spending decisions affect economic growth is a central issue to the discussion (Borner and Bodmer, 2004). It is reasonable to assume that budgetary rules play also a role for economic growth via the expectation channel signalling that the canton is committed to sound public finances.
9 However,
the empirical results do not support this view so far, probably because they have been introduced only recently in most cantons (for an overview see Stalder, 2005).
During the last 20 years the Swiss cantons have faced rather different developments. Zoug with 36.1% represents the canton with the highest rate of real economic growth per capita over the 1981–2001 period. With an economic growth per capita of �14% the development in Nidwalden is on the other extreme of the 26 cantons. On average, economic growth per capita accounted for 17% over the last 20 years on the cantonal level of Switzerland.
Figure 1 displays the maximum, minimum and average values of annual growth rates in Swiss cantons. A first look at the data on Fig. 2 reveals a slight negative correlation between the size of government and economic (R
2 ¼ 0.06). Thus, the
question occurs whether and how public expenditures
of Swiss cantons systematically affect the steady-state rate of economic growth as predicted by endogenous growth models (Romer, 1986; Barro, 1990; Barro and Sala-i-Martin, 1995; Mendoza et al., 1997).
In order to test the impact of the size of government on economic growth for rich jurisdic- tions, the methodology that has been applied is followed, for example by Kneller et al. (1999) or Fölster and Henrekson (2001). The empirical analysis is based on annual data for the 1981–2001 period for all 26 Swiss cantons. The dependent variable is cantonal GDP growth per capita and calendar year as calculated by BAK Basel Economics Ltd. The explanatory variables fall into three categories: (1) government expenditure per GDP as a proxy for government size and government expenditure per GDP split into public spending in the operating budget and public spending in the capital budget.
Fig. 1. Differentials of economic growth in Swiss cantons, 1981–2001
−4.0%
−3.0%
−2.0%
−1.0%
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
10% 15% 20% 25% 30% 35% 40%
Public expenditure in % of cantonal GDP
G ro
w th
o f re
a l G
D P
p e r
ca p ita
Fig. 2. Correlation between government size and economic
growth for 26 Swiss cantons, 5-year averages, 1981–2001
9 We are grateful to our referee for this argument.
1184 C. A. Schaltegger and B. Torgler
Current and consumption spending appear in the former budget while the latter budget consists of investment spending. (2) Initial GDP in order to incorporate the process of convergence as well as components of the production function: invest- ment, labour force and human capital, and (3) a set of socio-demographic indicators of the cantons as control variables. The regression equations have basically the following simplified form in logarithms:
yit � yit�1 ¼ �0 þ �1git þ �2 yit�1
þ �3Xit þ !i þ �t þ "it ð1Þ
where, yit is the log of GDP per capita in canton i of period t so that economic growth is described by yit � yit�1. The government size is specified by git which consists of the log of public expenditure per GDP. yit�1 on the right hand side of the equation incorporates the convergence process of economic growth between cantons. Xit is a vector of the differ- ent control variables of category (3) and the produc- tion function of category (2). Finally, there are three error components depicted by !i, �t and "it which represent state specific effects, year specific effects and the remaining error, respectively.
III. Results
Table 1 displays the basic results. As the tests on model selection (a Breusch-Pagan ML-test for selection between the OLS and the RE model and a Hausman-test for selection between the RE and the FE model) reveal, fixed effects estimates are most appropriate. Unlike Easterly and Rebelo (1993) but in line with Barro and Sala-i-Martin (1992), Agell et al. (1997, 1999) or Fölster and Henrekson (1999, 2001) initial GDP enters the regression with a highly significant negative coefficient. For the three conditioning variables, the investment ratio, labour force and higher schooling, there is no clear, empirically significant impact on economic growth for the Swiss cantons (coefficients are not statistically significant in all estimations).
The government size variable has statistically a significant negative coefficient, and the point estimate suggests that a decrease by one percentage point of GDP raises the growth rate by around 0.06 percentage points. Since the use of a short period of panel data may increase the risk that observed correlations are driven by business cycle effects, the unemployment rate is included in the regression as a control variable
that varies with the business cycle. Later, additionally five-year-averages will be used to tackle problems caused by business cycle effects (see Table 2). A typical business cycle correlation would imply that when growth rates fall government spending has to increase as a result of unemployment costs. Actually, it is assumed that this cyclical covariation is already moderated by controlling for period effects using time dummies. However, the highly significant and negative coefficients support the view that business cycles play an important role in explaining economic growth fluctuations.
10 The agglomeration variable
does not play a significant role in explaining econo- mic growth within Switzerland. This is somewhat surprising since it contradicts the notion that urban clusters play a prominent role in generating economic prosperity by spillover effects. However, it could be argued that there is not a perfect mapping of political borders with economic areas, which renders the agglomeration variables insignificant. The other socio-demographic factors represent control vari- ables to capture further state specific characteristics. This includes a language variable to control for systematic cultural differences according to the four official languages used in the 26 cantons.
The following applies some robustness tests of the above results. For example, Easterly and Rebelo (1993) argue in their article, that growth regression results are sensitive to the inclusion of the initial GDP. Therefore, in a variant of our basic regression, initial GDP is excluded.
However, even though the coefficient of the initial GDP is highly significant in the basic regression, Table 2 shows that an exclusion of this variable hardly changes the significance of the government size coefficient. Also, the negative quantitative effect decreases only slightly from �0.064 to �0.040. Second, the significant unemployment variable is excluded to show the effect of business cycle correlation on the other explanatory variables. Again and as indicated in Table 2, the exclusion of the unemployment rate hardly changes the results. Third, one has to address a fundamental identifica- tion problem due to the fact that there is a two-way causation between growth and government spending. While government spending affects growth by the supply side, growth itself affects government spend- ing by the demand side via income elasticity. The latter link refers to Wagner’s law arguing that there is an income elasticity over one for government services (Wagner, 1876). Chang (2002) provides
10 There is evidence for a positive relationship between government size and the unemployment rate (Christopoulos et al.,
2005). This points to a possible problem of multicollinearity in our regression. The findings of a simple correlation of 0.310 between these to variables do, however, not give rise to spurious results.
Growth effects of public expenditure 1185
evidence on Wagner’s law for Japan, the USA, the UK, South Korea, Taiwan and Thailand. The same applies for the G7 countries according to Kolluri et al. (2000) whereas Wahab (2004) finds only limited support for Wagner’s law when separating OECD countries with strong from those with weak economic growth.
We performed a Granger-causality test for the relation between public spending and growth. As Appendix C indicates government size in fact Granger-causes economic growth and not vice versa. This is in line with those results obtained by Ghali (1998) for ten OECD countries. However, even in the present case evidence was found for reversed timing of the changes of these two variables, it is not clear that this would have indicated reversed causality as Fölster and Henrekson (1999) point out. For example, long-term liabilities by the government as social security and pension schemes may reduce investment and hence growth long before expenditures
are actually increased. Thus, the timing of the spending and growth changes may not be a good indicator for causality. In any case, since the focus is on within-state variation by the fixed-effects estimates, the possible simultaneity bias is mitigated compared to cross-section analysis. This is because with yearly data, the likelihood that an error in the growth variable in one year will affect government spending in the same year is low.
However, while the use of panel data is reasonable in order to lower risks of simultaneity and to allow for within-state variation, there are also disadvan- tages of using annual data (Fölster and Henrekson, 2001). Estimating a panel of annual data without bias requires that the growth variable affects govern- ment spending in the same period, only, or to introduce the proper lag structure. Presumably, the proper lag varies between governments as well as over time so that they have not been captured accurately with yearly data. A solution to address these
Table 1. Regression results on the impact of government size on economic growth, 26 Swiss cantons, 1981–2001.
Dependent variable: per capita growth
Explanatory variables OLS FE OLS RE FE
Initial GDP p.c. �0.057*** �0.099*** �0.096*** �0.096*** �0.142*** (�3.00) (�5.66) (�4.86) (�5.39) (�7.27)
Government size �0.054*** �0.069*** �0.063*** �0.042*** �0.064*** (�3.64) (�5.02) (�3.82) (�3.06) (�4.53)
Investment �0.004 0.006 �0.005 �0.005 0.003 (�0.69) (1.07) (�0.74) (�0.85) (0.47)
Labour force �0.042* 0.023 �0.051** 0.019 0.046* (�1.71) (0.91) (�2.07) (0.71) (1.73)
Higher schooling 0.135*** �0.013 0.897*** 0.023 �0.021 (5.51) (�0.64) (3.41) (1.09) (�0.97)
Unemployment rate �0.002*** �0.001** �0.002*** �0.001*** �0.002*** (�4.62) (�2.39) (�4.37) (�2.69) (�3.56)
Agglomeration �0.008 0.004 0.012 (�0.38) (0.23) (0.76)
Population 0.045 �0.026 �0.109*** (1.20) (0.86) (�3.28)
Population > 65 0.335*** 0.183*** 0.129* (4.15) (2.69) (1.92)
Population < 15 �0.099 0.065 0.102 (�1.43) (1.02) (1.63)
German language 0.040 0.026 �0.031 (1.08) (0.87) (�0.96)
Canton effects Yes Yes Yes Yes Yes Year effects No Yes No Yes Yes LM test (Prob > Chi
2 ) 0.000
Hausman test (Prob > Chi 2 ) 0.000
Adjusted R-squared 0.116 0.622 0.168 0.197 0.638 No. of observations 546 546 546 546 546
Notes: t-statistics in parentheses. *, ** and *** denotes significance at the 10%, 5% and 1% level. For definitions of variables see Appendix A. For summary statistics see Appendix B. OLS: OLS estimates with canton dummies. RE: random effects estimates. FE: two-way fixed effects estimates.
1186 C. A. Schaltegger and B. Torgler
concerns is to focus on five-year averages. The results of the estimate using five-year-averages indicate that the level of significance for the impact of the size of government on economic growth drops to the 5% level, approximately. However, public spending still affects economic growth significantly negative.
Another way to address a possible simultaneity bias of government spending requires the estimation of instrumental variables (Barro, 1997). A common method is to use lagged values of the fiscal variables as instruments. However, in the fixed effects domain it is not possible to use lagged values. Kneller et al. (1999) and Fölster and Henrekson (2001) are there- fore followed and the regression estimated in first differences. The choice of instruments contains state dummy variables, lagged values of the government spending and initial GDP. The results of the instru- mental variable estimates are displayed in Table 2,
column 5. Again, the impact of public spending on economic growth is negative and significant. Anyhow, it has to be noted that even after introduc- ing instruments, the results of the coefficients may be biased. For example, Agell et al. (1999) are very sceptical about the instrumental variable technique in this case since the implemented instruments may still be correlated with the error term.
11 In a next
step we were looking for an instrument that is not correlated with the error term. Government central- ization is not significant in the first stage regression but is highly correlated with government size. Thus, government centralization serves as an appropriate instrument.
The results of the IV-regression in Table 2, column 6 indicate, that the negative correlation between government size and economic growth prevails even after the introduction of an alternative instrument. Comparing the results of Tables 1 and 2,
Table 2. Regression results on the impact of government size on economic growth for different specifications, 26 Swiss cantons,
1981–2001. Dependent variable: per capita growth
Explanatory variables FE FE-I FE-II AV FD-IV-I FD-IV-II
Initial GDP p.c. �0.142*** �0.132*** �0.022 �0.751*** �0.799*** (�7.27) (�6.73) (�0.85) (�9.35) (�8.12)
Government size �0.064*** �0.040*** �0.064*** �0.044** �0.174** �0.198* (�4.53) (�2.78) (�4.51) (�2.01) (�2.06) (�1.81)
Investment 0.003 �0.002 0.009 �0.003 0.002 0.003 (0.47) (�0.38) (1.42) (�0.25) (0.19) (0.35)
Labour force 0.046* 0.010 0.037 �0.084* 0.065** 0.069** (1.73) (0.37) (1.38) (1.91) (2.00) (2.04)
Higher schooling �0.021 �0.005 �0.029 �0.037 �0.015 �0.0133 (�0.97) (�0.22) (�1.31) (�0.88) (�0.76) (�0.65)
Unemployment rate �0.002*** �0.001** �0.002** �0.001 �0.001 (�3.56) (�2.37) (�2.23) (�0.54) (�0.40)
Agglomeration 0.012 0.001 0.013 0.001 0.007 0.005 (0.76) (0.06) (0.81) (0.05) (0.27) (0.17)
Population �0.109*** �0.016 �0.105*** 0.011 �0.549*** �0.556*** (�3.28) (�0.49) (�3.14) (0.28) (�4.41) (�4.45)
Population > 65 0.129* 0.069 0.051 0.0145* 0.261 0.249 (1.92) (0.98) (0.79) (1.71) (1.20) (1.14)
Population < 15 0.102 0.057 0.053 0.107 0.516** 0.514** (1.63) (0.86) (0.85) (1.38) (2.57) (2.54)
German language �0.031 �0.014 0.016 �0.063 �0.004 �0.010 (�0.96) (�0.40) (0.54) (�1.38) (�0.07) (�0.16)
Canton effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes R-squared 0.638 0.597 0.635 0.560 0.526 0.522 No. of observations 546 546 546 104 520 520
Notes: t-statistics in parentheses. * , ** and *** denotes significance at the 10%, 5% and 1% level. FE: two-way fixed effects estimates. AV: 5-year averages, two-way fixed effects. FD-IV-I: first differences with instrumental variable (government sizet�1). FD-IV-II: first differences with instrumental variable (government centralization). For definitions of variables see Appendix.
11 Agell et al. (1999, p. 363) write: ‘‘This procedure will, however, introduce more problems that it solves’’.
Growth effects of public expenditure 1187
it can be assumed that the effect of fiscal policy
decisions on economic growth is not simply due to
simultaneity and endogeneity. Inspite of different
econometric reasons to be careful with the interpreta-
tion of the estimated coefficients there is at least some evidence supporting the view that the expansion of
the size of government hampers economic growth in a
rich country with a developed public sector. The
estimated results indicate a significant and negative
correlation between public spending and the rate of economic growth throughout the different specifica-
tions and estimation techniques. While the obtained
results may be interpreted as evidence for a crowding-
out effect, where public expenditure displace private
sector productivity, it is reasonable to assume that
investment spending have a different impact on economic growth than transfer spending or public
consumption (Barro, 1990). The traditional approach
is to divide public spending into the two broad
categories of public consumption and public invest-
ment. The former is said to retard economic prosper- ity while the latter should promote growth prospects.
Implicitly, consumption spending are classified
as unproductive and growth-retarding public pro-
grammes whereas investments fall into the category
of productive and growth-inducing government
activities. This distinction has an intuitive appeal but is also problematic since investment projects
can be wasteful as well while public consumption
need not necessarily be unproductive (Tanzi and
Zee, 1997). Hence, some authors distinguish between produc-
tive and unproductive government activities by sorting all spending tasks of the budget according to that
criterion. For example, Kneller et al. (1999) show that
productive government expenditure enhance growth
rates of the 22 OECD countries over the 1970–1995
period analysed in their empirical study. In order to get a more detailed picture of different
public spending impacts on growth for the Swiss
sub-federal governments, we distinguish between
spending from the capital budgets to finance invest-
ments and spending in the operating budget to
finance current expenditure. 12
Figures 3 and 4 give
a first indication that growth impacts of the two budgets are different. While the picture for current
spending is very similar to the picture for overall
government spending, there is no correlation between
investment spending and economic growth. The first
clue becomes confirmed in the multivariate analysis.
As can be seen by Table 3, there is evidence that the growth impact differs between spending from the
current budget and spending from capital budgets. While current spending has a significantly strong
negative impact on economic prosperity of cantons, the same does not hold for investment spending.
This result confirms that the type of government
expenditure matters for economic growth. Thus, the results are very much in line with those obtained
by Romero de Ávila and Strauch (2003) for the
−4.0%
−3.0%
−2.0%
−1.0%
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
10% 15% 20% 25% 30%
Public expenditure in % of cantonal GDP
G ro
w th
o f re
a l G
D P
p e r
ca p ita
Fig. 3. Correlation between current public spending and
economic growth, 26 Swiss cantons, 1981–2001, 5-year
averages
−4.0%
−3.0%
−2.0%
−1.0%
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
0% 2% 4% 6% 8% 10% 12% 14%
Public expenditure in % of cantonal GDP
G ro
w th
o f re
a l G
D P
p e r
ca p ita
Fig. 4. Correlation between current public investment
spending and economic growth, 26 Swiss cantons,
1981–2001, 5-year averages
12 However, as argued by Shepsle and Weingast (1984) it is possible that such a distinction of the budgets only affects the
labelling of government spending without affecting the composition of spending. For example, cantons with fiscal requirements for the operating budgets may try to relabel operating expenditure as capital projects in order to evade constitutional spending limitations. Poterba (1995) finds empirical evidence for the US states that states with separate capital budgets spend more on public capital projects than comparable states with unified budgets.
1188 C. A. Schaltegger and B. Torgler
European countries. Both studies find that govern- ment consumption negatively affect growth rates of GDP per capita, while public investment has a positive impact.
IV. Conclusions
There is a huge empirical literature investigating the relationship between government size and eco- nomic growth. To date, the cross-country empirical evidence on growth effects of public expenditure is still inconclusive, however. Theoretically, this is not surprising since small or big government by itself is not an asset. A negative relationship should only apply for rich countries with a large public sector while in developing countries a growing size of government typically reveals safer property rights and
the enforcement of contracts. Thus, there is no reason to believe that small governments will generally promote economic growth. In this respect, restricting the analysis on rich countries only may give one a more detailed picture of the issue. Naturally, a selection of a sub-sample of rich countries is always somewhat arbitrary. Another possibility is to concentrate on governments within a rich country. However, only few studies investigate the effect of state and local spending on economic growth. This study attempted to test the impact of the size of government on economic growth for the sub-federal level of a rich country using panel data of a full sample of the 26 Swiss cantons over the 1981–2001 period. The general finding is a fairly robust negative relationship between government size and economic growth. Even though we do not claim to settle the issue, the results are found to be
Table 3. Regression results on the impact of government size on economic growth distinguishing for different budgets,
26 Swiss cantons, 1981–2001. Dependent variable: per capita growth
Explanatory variables FE AV FD-IV-I FD-IV-II
Initial GDP p.c. �0. 239*** �0.047 �0.710*** �0.916*** (�10.65) (�1.27) (�9.11) (�7.51)
Government size I (current spending) �0.148*** �0.047* �0.364*** �0.558*** (�8.89) (�1.76) (�4.62) (�4.54)
Government size II (investment spending) 0.006 �0.003 0.015 0.009 (1.62) (�0.34) (0.80) (0.44)
Investment �0.007 �0.009 �0.005 �0.001 (�1.21) (�0.74) (�0.78) (�0.06)
Labour force 0.038 �0.092** 0.060* 0.075* (1.51) (2.13) (1.77) (1.86)
Higher schooling �0.021 �0.035 0.007 0.029 (�1.02) (�0.82) (0.31) (1.07)
Unemployment rate �0.003*** �0.003** �0.001 �0.001 (�5.09) (�2.41) (�0.79) (�0.21)
Agglomeration 0.006 0.003 0.013 �0.001 (0.39) (0.14) (0.46) (�0.02)
Population �0.117*** 0.003 �0.376*** �0.308* (3.72) (0.08) (�2.72) (�1.86)
Population > 65 0.018 0.141 0.070 �0.053 (0.28) (1.62) (0.30) (�0.19)
Population < 15 �0.001 0.077 0.356* 0.272 (�0.02) (0.98) (1.70) (1.09)
German language �0.031 �0.063 �0.046 �0.090 (�1.02) (�1.35) (�0.71) (�1.15)
Canton effects Yes Yes Yes Yes Year effects Yes Yes Yes Yes R-squared 0.671 0.560 0.485 0.286 No. of observations 546 104 520 520
Notes: t-statistics in parentheses. * , ** and *** denote significance at the 10%, 5% and 1% level. FE: two-way fixed effects estimates. AV: 5-year averages, two-way fixed effects. FD-IV-I: first differences with instrumental variable (government sizet�1). FD-IV-II: first differences with instrumental variable (government centralization). For definitions of variables see Appendix.
Growth effects of public expenditure 1189
robust also after adopting changes in specification and applying different estimation techniques.
Anyhow, theory does not only predict that fiscal policy affects growth by the level of government spending but also by the expenditure structure. That’s why we test the effect of government spending of the operational budget separately from the impact of investment spending from the capital budget. Consistent with Barro’s (1990) predictions, an increase in public spending from operating budgets significantly reduces growth while there is no significant impact on economic growth by expenditure from capital budgets.
Acknowledgements
The authors would like to thank BAK Basel Economics for the provision of data for cantonal GDPs and René L. Frey, Lars P. Feld, Bruno Jeitziner, Gebhard Kirchgässner and Jordan Rappaport, the editor Lucio Sarno and an anon- ymous referee for helpful comments. All remaining errors are of course the authors’ responsibility.
References
Agell, J., Lindh, T. and Ohlsson, H. (1997) Growth and the public sector: a critical review essay, European Journal of Political Economy, 13, 33–52.
Agell, J., Lindh, T. and Ohlsson, H. (1999) Growth and the public sector: a reply, European Journal of Political Economy, 15, 359–66.
Agell, J., Ohlsson, H. and Thoursie, P. S. (2003) Growth effects of government expenditure and taxation in rich countries: a comment, Working Paper University of Stockholm.
Alesina, A., Ardagna, S., Perotti, R. and Schiantarelli, F. (2002) Fiscal policy, profits, and investment, American Economic Review, 92, 571–89.
Atkinson, A. B. (1995) The welfare state and economic performance, National Tax Journal, 47, 171–98.
Barro, R. J. (1990) Government spending in a simple model of endogenous growth, Journal of Political Economy, 98, 103–25.
Barro, R. J. (1991) Economic growth in a cross section of countries, Quarterly Journal of Economics, 106, 407–43.
Barro, R. J. (1997) Determinants of Economic Growth: A Cross-Country Empirical Study, MIT Press, Cambridge, MA.
Barro, R. J. and Sala-i-Martin, X. (1992) Convergence, Journal of Political Economy, 100, 223–51.
Barro, R. J. and Sala-i-Martin, X. (1995) Economic Growth, McGraw-Hill, New York.
Bernholz, P. (2004) Der Wettbewerb der Systeme: Arten, Interaktionen und Folgen, in Perspektiven der Wirtschaftspolitik (Eds) C. A. Schaltegger and S. C. Schaltegger, vdf, Zürich, pp. 241–58.
Borner, S. and Bodmer, F. (2004) Wohlstand ohne Wachstum – Eine Schweizer Illusion, Orell Füssli, Avenir Suisse.
Chang, T. (2002) An econometric test of Wagner’s law for six countries based on cointegration and error- correction modelling techniques, Applied Economics, 34, 1157–69.
Christopoulos, D. K., Loizides, J. and Tsionas, E. G. (2005) The Abrams curve of government size and unemploy- ment: evidence from panel data, Applied Economics, 37, 1193–9.
de Soto, H. (2002) Law and property outside the west: a few new ideas about fighting poverty, Forum for Development Studies, 29, 349–61.
Easterly, W. and Rebelo, S. (1993) Fiscal policy and economic growth: an empirical investigation, Journal of Monetary Economics, 32, 417–58.
Engen, E. M. and Skinner, J. (1992) Fiscal policy and economic growth, NBER Working Paper 4223.
Feld, L. P., Kirchgässner, G. and Schaltegger, C. A. (2003) Decentralized taxation and the size of government, evidence from Swiss sub-federal governments, CESifo Working Paper No. 1087.
Fölster, S. and Henrekson, M. (1999) Growth and the public sector: a critique of the critics, European Journal of Political Economy, 15, 337–58.
Fölster, S. and Henrekson, M. (2001) Growth effects of government expenditure and taxation in rich countries, European Economic Review, 45, 1501–20.
Ghali, K. H. (1998) Government size and economic growth: evidence from a multivariate cointegration analysis, Applied Economics, 31, 975–87.
Grier, K. B. (1997) Governments, unions and economic growth, in Government and Growth (Ed.) V. Bergström, Clarendon Press, Oxford, pp. 149–201.
Gwartney, J., Lawson, R. and Holcombe, R. (1998) The size and functions of government and economic growth, paper prepared for the US Joint Economic Committee.
Hansson, P. and Henrekson, M. (1994) A new framework for testing the effect of government spending on growth and productivity, Public Choice, 81, 381–401.
Helms, L. J. (1985) The effect of state and local taxes on economic growth: a time series-cross section approach, Review of Economics and Statistics, 67, 574–82.
Holcombe, R. G. and Lacombe, D. J. (2004) The effect of state income taxation on per capita income growth, Public Finance Review, 32, 292–312.
Hsieh, E. and Lai, K. S. (1994) Government spending and economic growth: the G-7 experience, Applied Economics, 26, 535–42.
Islam, M. Q. (2004) The long run relationship between openess and government size: evidence from Bounds test, Applied Economics, 36, 995–1000.
Keefer, P. and Knack, S. (1997) Why don’t poor countries catch up?, A cross-national test of an institutional explanation, Economic Inquiry, 35, 590–602.
Kneller, R., Bleaney, M. F. and Gemell, N. (1999) Fiscal policy and growth: evidence from OECD countries, Journal of Public Economics, 74, 171–90.
Kolluri, B. R., Panik, M. J. and Wahab, M. S. (2000) Government expenditure and economic growth: evidence from G7 countries, Applied Economics, 32, 1059–68.
1190 C. A. Schaltegger and B. Torgler
Mendoza, E. G., Milesi-Ferretti, G. and Aseac, P. (1997) On the ineffectiveness of tax policy in altering long-run growth: Harberger’s superneutrality conjecture, Journal of Public Economics, 66, 99–126.
Poterba, J. M. (1995) Capital budgets, borrowing rules, and state capital spending, Journal of Public Economics, 56, 165–87.
Rappaport, J. (1999) Local growth empirics, CID-Harvard University Working Paper No. 23.
Romer, P. M. (1986) Increasing returns and long-run growth, Journal of Political Economy, 99, 500–21.
Romero de Ávila, D. and Strauch, R. (2003) Public finances and long-term growth in Europe – evidence from a panel data analysis, ECB Working Paper No. 246.
Schaltegger, C. A. (2002) Budgetregeln und ihre Wirkung auf die öffentlichen Haushalte: Empirische Ergebnisse aus den US-Bundesstaaten und den Schweizer Kantonen, Schmollers Jahrbuch, 122, 369–413.
Shepsle, K. A. and Weingast, B. R. (1984) Legislative politics and budget outcomes, in Federal Budget Policy in the 1980s (Eds) G. Mills and J. Palmer, The Urban Institute Press, Washington, D.C. pp. 343–67.
Singh, R. J. and Weber, R. (1997) The composition of public expenditure and economic growth: can
anything be learned from Swiss data?, Swiss Journal of Economics and Statistics, 133, 617–34.
Slemrod, J. (1995) What do cross-country studies teach about government involvement, prosperity, and economic growth?, Brookings Papers on Economic Activity, 2, 373–431.
Slemrod, J. (1998) How costly is a large, redistributive public sector?, Swedish Economic Policy Review, 5, 87–105.
Stalder, K. (2005) Fiscal rules for cantons and communes, Mimeo, University of St Gallen.
Tanzi, V. and Zee, H. H. (1997) Fiscal policy and long-run growth, IMF Staff Papers, 44, 179–209.
Tanzi V. and Schuknecht, L. (2000) Public Spending in the 20th Century, Cambridge University Press, Cambridge.
Vedder, R. K. and Gallaway, L. E. (1998) Government size and economic growth, Paper prepared for the US Joint Economic Committee.
Wagner, A. (1876) Grundlegung der politischen Ökonomie, Winter’sche Verlagshandlung, Leipzig.
Wahab, M. (2004) Economic growth and government expenditure: evidence from a new test specification, Applied Economics, 36, 2125–35.
Weingast, B. R. (1995) The economic role of political institutions: market-preserving federalism and economic development, Journal of Law, Economics, and Organization, 11, 1–31.
Appendix A
Data description
Variable name Description Source
Government size Total cantonal and local expenditure per GDP Own calculations on the basis of Swiss Federal Finance Administration and BAK Basel Economics
Current spending Cantonal and local expenditure in the operational budget per GDP
Own calculations on the basis of Swiss Federal Finance Administration and BAK Basel Economics
Investment spending Cantonal and local expenditure in the capital budget per GDP
Own calculations on the basis of Swiss Federal Finance Administration and BAK Basel Economics
GDP Real cantonal GDP BAK Basel Economics Investment Cantonal and local investment spending
per GDP Swiss Federal Statistical Office
Labor force Share of employment on the cantonal population
Swiss Federal Statistical Office
Higher schooling Share of population with secondary education on the cantonal population
Swiss Federal Statistical Office
Unemployment rate Share of unemployment on the cantonal population
Own calculations on the basis of Swiss Federal Statistical Office
Agglomeration Proportion of local communities having more than 10 000 inhabitants.
Swiss Federal Statistical Office
Population Cantonal population Swiss Federal Statistical Office Population >65 Share of cantonal population over the age 65
on total cantonal population Swiss Federal Statistical Office
Population <15 Share of cantonal population under the age 15 on total cantonal population
Swiss Federal Statistical Office
German language Dummy ¼ 1 for German speaking cantons Own investigations Government centralization Share of cantonal public spending on cantonal
and local spending Swiss Federal Finance Administration
Growth effects of public expenditure 1191
Appendix B
Descriptive statistics
Variable Mean Std. dev. Minimum Maximum
Government size 0.226 0.047 0.118 0.386 Current spending 0.183 0.037 0.098 0.291 Investment spending 0.044 0.023 0.012 0.145 GDP 41 590 13 064 26 324 117 228 Investment 0.160 0.055 0.050 0.477 Labor force 0.480 0.032 0.396 0.564 Higher schooling 0.137 0.059 0.023 0.334 Unemployment rate 0.018 0.018 0 0.078 Agglomeration 0.324 0.249 0 0.995 Population 261 938 272 497 12 781 1 228 628 Population >65 0.146 0.021 0.103 0.210 Population <15 0.186 0.024 0.113 0.241 German language 0.714 0.353 0.050 0.980 Government centralization 0.673 0.106 0.510 0.990
Notes: For a detailed description of the variables see Appendix A. All statistics are computed for 546 observations.
Appendix C: Granger Causality Test
Part 1. Did the Public Expenditures Come First? The following equation was estimated by OLS:
ðyt � yt�1Þt ¼ �0 þ �1gt�1 þ �2ðyt � yt�1Þt�1 þ "t
H0: Public expenditures do not Granger cause economic growth No. of lags P > |t| Adj. R-squared 1 0.019 0.034
Part 2. Did the Economic Growth Come First? The following equation was estimated by OLS:
gt¼�0 þ �1gt�1 þ �2ðyt � yt�1Þt�1 þ "t
H0: Economic growth does not Granger cause public expenditures No. of lags P>|t| Adj. R-squared 1 0.756 0.102
If lagged values of public expenditures help predict current values of growth (case 1) in a forecast formed from lagged values of growth and public expenditures, then public expenditures are said to Granger cause growth (see part 1). A symmetric regression tests the reverse causality. The Granger causality test is performed using one lags and working with five-year-averages to reduce possible business cycle effects. Appendix C allows one to identify which came first, rejecting the non-causality of the impact of growth on public expenditures and at the same time failing to reject the non-causality of the impact of public expenditures on growth (statistically significant negative coefficient). Therefore, we can conclude that public expenditures came first.
1192 C. A. Schaltegger and B. Torgler