Discretionary Grants and Soft Budget Constraints in Switzerland
By
Lars Feld
University of Marburg
feld@wiwi.uni-marburg.de
and
Timothy J. Goodspeed
Hunter College and CUNY Graduate Center
timothy.goodspeed@hunter.cuny.edu
March 17, 2005
Abstract: Theoretical models of soft budget constraints in a federation suggest a dynamic
relationship between the spending and deficit patterns of lower levels of government on
the one hand, and central government grant decisions on the other. Central governments
may feel the obligation to increase future discretionary grants to profligate lower levels of
government (even if they promise not to). This lowers the lower level government’s
expected opportunity cost of spending and borrowing, and increases the incentive of
lower-level governments to be profligate currently. Moreover, this pattern should be
more pronounced the greater is the discretionary power of the central government to vary
grants. We use data on 26 Swiss cantons over the period 1980 – 2001 to test this
hypothesis.
I. Introduction
The move toward decentralized government around the world, and the troubles
that have been encountered, has spawned research in new areas of fiscal federalism. The
problems in implementing intergovernmental grants, and the study of soft budget
constraints in general, is one of these areas. Wildasin (1997) is one of the first papers to
explore these issues; he emphasizes the externalities that can result from governmental
bankruptcies and how these externalities may be positively related to the size of the
failing entity. Inman (2001) pursues the idea that some central governments may develop
a “hard budget” reputation and thereby deter profligate behavior of lower level
governments. Goodspeed (2002) points to the dynamic interaction of central and lower
level governments, showing that forces predicted by lower level governments can lead
the central government astray, with lower level governments taking advantage of future
central government bail-out policies.
In all of these models, the discretionary power of the central government is key.
If the central government can use discretionary power to change grant allocations in the
future, it will be tempting to bail-out lower levels of government. On the other hand, if
the central government ties its hands by using grants that are non-discretionary and
cannot be changed in the future, the central government will force itself to implement a
hard budget constraint policy. This suggests that one should observe more profligate
behavior when lower level governments receive more discretionary grants. The primary
purpose of this paper is to test this proposition using data on Switzerland. We develop a
simple dynamic econometric approach to do this.
Some papers such as Buettner and Wildasin (2004?), Jones, Sanguinetti, and
Tommasi (1999), Borge and Rattso (1999), Von Hagen and Dahlberg (2002), and GarciaMila, Goodspeed, and McGuire (2001) have attempted to estimate certain aspects of soft
budget constraints. Our econometric approach differs in two important ways from these
papers. First, we explicitly consider the dynamic relationship between the central and
lower levels of government. Second, we exploit the fact that discretionary grants are
more prone to manipulation than non-discretionary grants.
Switzerland is an interesting country to study for several reasons. It is of course a
highly decentralized and federal country that uses grants extensively. Moreover, as
discussed later, its grant system encompasses both discretionary and non-discretionary
grants that vary over time as well as across cantons. Our preliminary results suggest that
more discretionary grants lead to greater spending and debt levels, though we do not find
evidence that they lead to greater deficits over our sample. Greater non-discretionary
grants lead to substantially lower debt levels, with weaker evidence that they lead to
lower deficits.
Our paper is organized as follows. We develop a simple theoretical model in the
next section that illustrates the difference between discretionary and nondiscretionary
grants. In section 3, we discuss grants in Switzerland. We develop an econometric
model that tries to capture the simple dynamics between the central government and
cantons in section 4, and also present our preliminary results there. Section 5 concludes.
II. A Basic Model
To illustrate the basic difference between discretionary and non-discretionary
grants, consider the following simple two-period model, based on Goodspeed (2002).
The federal economy consists of two types of players, regional governments and the
central government. Region i is inhabited by ni identical people whose utility is assumed
to be a function of private consumption in periods 1 and 2, Ci1 and Ci2, and per-capita
public consumption in periods 1 and 2, Gi1 and Gi2. The representative consumer has
private income in each period, Yi1 and Yi2. Before play begins, the central government is
assumed to decide on an initial level of grants for each region in period one, denoted gi1.
This initial decision is exogenous to the game to be played. In addition to the exogenous
central government grants received in period one, the region is able to borrow an amount
per capita for public consumption in period 1 denoted Bgi1, and chooses a period one tax
rate. Consumers can borrow an amount for private consumption in period 1 denoted by
Bci1. In period two, regions choose a tax rate and the central government chooses second
period per-capita grants, gi2.
There are strategic interactions among the regions, and between the regions and
the central government. Consider fist the interaction between regional governments.
Regional governments are assumed to move simultaneously and make a choice
concerning borrowing and taxation in period one. The equilibrium is a Nash equilibrium
in which each region takes the other regions’ borrowing and tax rates as given. In period
two, regional governments again move simultaneously and choose a period two tax rate.
The central government moves in period two and chooses regions’ second period
grant level. With respect to regional borrowing and first period taxation, the interaction
between the central government and regional governments is therefore sequential with
the regional government moving first. The game is of the Stackelberg variety, that is one
of perfect information, so that the regional government knows how the central
government will react in period two, and takes into account the reaction function of the
central government in its choice of borrowing in period one.
We consider two different possibilities for the central government distribution of
period two grants. In a first possibility, the central government is assumed to follow a
rule so that grants are non-discretionary. Such non-discretionary grants will be defined as
grants that cannot be changed by the central government. In other words, the central
government is committed to a rule to distribute grants in a particular way in period 2. In
a second possibility, the central government will be assumed to have some discretion
over the distribution of grants. It will be able to choose to increase future grants to a
region if it so chooses. As we will see, this discretion can lead to a soft budget constraint,
and hence greater borrowing or lower local taxes than would occur with nondiscretionary grants.
The non-discretionary grant case. Under non-discretionary grants, the central
government is committed to a rule under which regional grants in period two are
unchanging regardless of the choices made by the regional government in period one.
The central government decision on second period grants is thus quite simple:
g *j 2 = Cons tan t
Given this rule for period two grants, regional governments decide how much to borrow
and tax in period one. Regional government i’s problem is
Max
G
BiC1 i , Bi1 , t i1, t i 2
u i ( G i1 ) + vi ( Gi 2 ) + wi ( C i1 ) + z i ( C i 2 )
s.t. Gi1 = g i1 + t i1 Y i1 + BiG1
C
C i1 = Y i1 (1 - t i1 ) + Bi1
*
G
G i 2 = g i 2 + t i 2 Y i 2 - Bi1 (1 + r)
C
c
C i 2 = Y i 2 (1 - t i 2 - t 2 ) - Bi1 (1 + r)
t2 ∑ n j Y j2 = ∑ n j g j2
*
c
j
j
*
j2
g = Cons tan t(1 + r) for all j
Given the nondiscretionary nature of central government grants, the regional
government’s incentives are efficient, both intertemporally and contemporaneously
between public and private goods. This is easily seen form the regional government’s
first order conditions:
∂ ui
∂ wi
Y i1 Y i1 = 0
∂ G i1
∂ C i1
∂ wi
∂ zi
- (1 + r)
=0
∂ C i1
∂C i 2
∂ u i ∂ vi
(1 + r) - = 0
∂ G i1 ∂ G i 2
∂ vi
∂ zi
Y i2 Y i2 = 0
∂ Gi2
∂C i 2
The discretionary grant case. Under discretionary grants, the central government
changes a regional government’s second period grants depending on circumstances.
Theoretically, the exact way in which discretionary grants will be disbursed depends on
how the central government is modeled. A simple model of central government
behavior, derived in Goodspeed (2002), supposes that second period grants are chosen so
as to maximize expected votes:
Max ∑ ni pi ( u i ( G i1 ) + vi ( Gi 2 ) + wi ( C i1 ) + z i ( C i 2 ))
g2
i
s.t. Gi1 = g i1 + t i1 Y i1 + BiG1
C
C i1 = Y i1 (1 - t i1 ) + Bi1
*
G
G i 2 = g i 2 + t i 2 Y i 2 - Bi1 (1 + r)
C
c
C i 2 = Y i 2 (1 - t i 2 - t 2 ) - Bi1 (1 + r)
t2 ∑ n j Y j2 = ∑ n j g j2
*
c
j
j
The first order conditions simplify to:
∂ pi ∂ vi ∂ p j ∂ v j
=
for all i, j
∂ vi ∂ G i 2 ∂ v j ∂ G j 2
The central government equates the weighted marginal utility of regions’ voters where
the weights are the change in the probability of voting for the incumbent. These first
order conditions implicitly define the central government’s discretionary allocation of
second period grants. As discussed in Goodspeed (2002), regional government
borrowing in this model will induce changes in discretionary grants from the central
government for two reasons. First, the regional government is effectively forcing the
hand of the central government by purposefully lowering its second-period public good
consumption. The central government responds by increasing grants to maintain its
desired public consumption level for region i in period 2. Second, the regional
government can effectively increase the political payoff of any central government
increase in grants by borrowing. For both of these reasons, increased borrowing in period
one by the regional government will induce more discretionary grants in period two from
the central government.
Moreover, the regional government knows this and consequently has an incentive
to overspend in period one and try to induce more discretionary grants in period two.
This can be seen by examining the regional government’s problem, noting that the
regional government now expects to receive more discretionary grants when it borrows:
Max
G
BiC1 i , Bi1 , t i1, t i 2
u i ( G i1 ) + vi ( Gi 2 ) + wi ( C i1 ) + z i ( C i 2 )
s.t. Gi1 = g i1 + t i1 Y i1 + BiG1
C
C i1 = Y i1 (1 - t i1 ) + Bi1
*
G
G i 2 = g i 2 + t i 2 Y i 2 - Bi1 (1 + r)
C
c
C i 2 = Y i 2 (1 - t i 2 - t 2 ) - Bi1 (1 + r)
t2 ∑ n j Y j2 = ∑ n j g j2
*
c
j
j
∂p
 ∂

g i*2 = f i  vi ( Bi1 ) , i ( Bi1 ) (1 + r)
∂ vi
 ∂ Gi 2

The last constraint is regional government i’s expectation of its future grant allocation.
Notice that the regional government recognizes the discretionary nature of the grant, and
predicts that it will receive more grants in period 2 as it borrows in period 1. The first
order conditions are:
∂ ui
∂ wi
Y i1 Y i1 = 0
∂ G i1
∂ C i1
∂ wi
∂ zi
- (1 + r)
=0
∂ C i1
∂C i 2


∂ u i  ∂ f i 2  ∂ vi
∂ zi 

- 1 (1 + r) ∂ G i1  ∂BiG1  ∂ Gi 2
∂C i 2 



ni Y i 2
m
∑n Y
j
j2
∂ f i2
∂ BiG1
j=1
∂ vi
∂ zi
Y i2 Y i2 = 0
∂ Gi 2
∂C i 2
The first order conditions simplify to:
∂ ui
∂ Gi1
=
∂ vi
∂Gi 2

 ni Y i 2 ∂ f
i2

G
 ∑ n j Y j 2 ∂ Bi1
 j


 ∂f 
(1 + r) +  1 - Gi 2 (1 + r)

 ∂Bi1 




(1 + r)  = 0



The first term represents the additional cost to a region that results from its increased
allocation of grants. This does not represent the full cost since the additional costs of $1
of additional grants are shared among all regions. Given the assumption of a proportional
income tax, the costs are shared in proportion to the total income of the region. The
second term represents the reduced opportunity cost of borrowing that results from the
central government increasing grants to region i. Effectively, the central government
bails out region i by increasing its grant allocation.
3.
Intergovernmental grants and institutional variation
The predictions of the model are tested using data for Switzerland. Swiss
federalism is characterized by a far reaching fiscal autonomy at the state and local levels.
The states (cantons) have the basic power to tax personal and corporate income as well as
capital while the local jurisdictions levy a surcharge on the cantonal income tax and raise
own wealth and property taxes. The federal government mainly relies on indirect
(proportional) taxes but also on a highly progressive income tax. In addition, it levies a
source tax on capital income by a uniform rate of 35 percent that could be deducted by
taxpayers when declaring their income to the cantonal tax administration.
The Swiss system of intergovernmental grants on the federal level has been
established in 1959. Today, it consists of a complex system of almost exclusively
vertical transfers from the federal government to the 26 cantons. There is no horizontal
equalization system in Switzerland, but only specific horizontal inter-cantonal payments
for particular services. However, the vertical transfers have a strong horizontally
equalizing impact since rich cantons bear most of the financial burden of the system. The
system of intergovernmental grants has four main ‘pillars’: The federal contributions to
the cantons, the cantonal share of federal revenues, the cantonal share of the gains from
the central bank and cantonal contributions to the federal social security system. 30% of
the highly progressive federal income tax is directly redistributed to the cantons
according to income (17%) and to financial prosperity of the canton (13%). The federal
source tax on capital income is distributed in a similar fashion to the cantons, but instead
of financial prosperity the population is taken into account. Both grants are lump-sum
and only the first component is not following fiscal equalization criteria. The remaining
70% of the revenue from the federal income tax are used to fund matching grants.
Table 1: Federal index of financial prosperity and federal index for mountainous areas (for 2002/2003),
share of grants in the total cantonal revenues and share of conditional grants, 1999
Cantons
Share of grants
Federal index of Federal index of
Share of
from total canfinancial prosmountainous
matching grants
tonal revenues
perity (Swiss
areas (Swiss
(%)
(%)
average = 100) average = 100)
Financially potential cantons
Zug
26.1
38.4
216
96.70
Basel-City
10.8
71.2
173
111.03
Zurich
15.1
66.5
160
108.95
Geneva
Nidwalden
9.7
39.6
61.1
82.6
141
129
111.08
84.12
Basel-Land
15.0
71.4
120
105.37
Cantons with average
financial potential
Schwyz
40.4
51.3
112
85.67
Schaffhausen
Aargau
17.7
19.3
69.8
70.8
107
97
111.03
110.43
Vaud
19.1
77.4
94
106.22
Thurgau
25.7
77.3
83
110.40
Solothurn
Glarus
26.8
26.8
77.8
59.0
82
82
103.42
77.16
Ticino
23.3
74.1
82
86.06
St. Gallen
24.9
79.5
80
98.73
Graubünden
Luzern
47.1
27.9
87.7
77.0
77
67
70.00
102.12
Uri
48.8
89.7
64
73.55
Appenzell a.Rh.
29.6
69.8
63
82.03
Appenzell i.Rh.
38.7
80.1
62
71.33
Bern
28.2
78.9
57
94.05
Neuchâtel
38.8
81.7
55
88.56
Fribourg
35.3
80.1
51
97.09
Obwalden
44.5
79.3
35
76.83
Jura
48.6
84.8
34
84.98
Valais
41.7
72.8
30
81.4
Switzerland
23.1
74.9
100
100
Financially weak cantons
Source: Swiss Federal Finance Department 1999. Notes: The federal index of financial prosperity consists
of four sub-indices. The federal transfer index for mountainous areas is one out of four. The lower the
number of the index the higher the amount of transfers a canton receives.
The share of grants from total cantonal revenue is reported in the second column
of Table 1 where cantons are listed according to their financial strength. In some of the
financially weak cantons, for example Uri and Jura, grants account for almost half of the
total cantonal income. Other cantons like Zurich, Geneva or both Basel fund their
spending to more than 85 percent by own taxes. On average, grants cover less than a
quarter of total cantonal revenue. As Table 1 also indicates, on average three quarters of
all federal transfers take the form of matching grants. Again there is an interesting
variation ranging from 38 percent in the case of Zug to almost 90 percent in the case of
Uri. Zug appears to be an outlier since the share of matching grants is at least 50 percent
in the other cantons. Comparing the descriptive figures for Zug it becomes obvious that
this canton is obtaining a relatively large amount of lump-sum grants from the federal
government.
The amount of matching grants received by a canton depends on an index which
should reflect cantonal prosperity. This index relies on four components: cantonal
income, cantonal tax revenues adjusted by the cantonal tax burden, the cantonal tax
burden itself and an index reflecting the share of mountainous areas in a canton. The
index of financial prosperity and its sub-index of mountainous areas is presented in the
two last columns in Table 1. The index of financial prosperity determines the order of
cantons in the table. The Swiss average is set to 100 index points, while the deviations
from the average determine the range of the index. It ranges from 216 in Zug to 30 index
points in the canton of Valais. The financial strength of Zug is more than 7 times that of
the Valais. It is also obvious from Table 1 that neither financial prosperity nor the
variance of it can be fully explained by the share of mountainous area in a canton. This
latter range is much lower and the richer half of the cantons appears to be assessed as
having a lower share of mountains.
4. The empirical model
The above model suggests that the impact of discretionary grants can be estimated
via a simple two-equation dynamic empirical model. The discretionary nature of the
grants suggests that there are two behavioral relationships to estimate, the central
government grant decision and the regional government spending and borrowing
decision. According to the theory, the regional government will decide on period one
spending and borrowing, using a prediction of how the central government will react in
the future. Predicting that the central government will increase grants in the future leads
to present period profligate behavior. In period two, the central government uses its
discretion to change the grant allocation of region i as region i has expected. In contrast,
for the non-discretionary grant case, the central government does not respond in period
two, forcing the region to confront a hard budget constraint and non-profligate spending
incentives in the present. Hence, one would expect that a greater expected (future)
amount of discretionary grants will lead to greater spending and borrowing by regional
governments.
This two equation dynamic model could additionally be split by distinguishing the
spending and borrowing decision of regional governments. Furthermore, borrowing is
captured by budget deficits, which capture short-term borrowing, and public debt, which
reflects long-term borrowing behavior. Thus, the following econometric model is
formulated:
yit = α0 + α1 Matchingit+2 + α2 Lump-Sumit + α3 Incit + α4 Popit + α5 Urbanit
+ α6 Oldit + α7 Youngit + α8 Unemplit + ε
Matchingit+2 = β0 + β1 Incit+2 + β2 Popit+2 + β3 Urbanit+2 + β4 Oldit+2
+ β5 Youngit+2 + β6 Unemplit+2 + β7 Mountainit+2 + µ
where the dependent variable y, stands for the following variables (all in per capita):
cantonal expenditure, cantonal debt and the cantonal deficit. The explanatory variables
are:
Matching federal discretionary grants (matching grants) per capita,
Lump-Sum federal unconditional grants per capita,
Inc
disposable income per capita,
Pop
population,
Urban
share of urban population,
Old
share of population older than 65,
Young
share of population younger than 20,
Mountain federal index of mountainous areas,
ε, µ
stochastic terms.
In addition to matching grants, the fiscal policy models contain standard control
variables like income, population size and population structure (urban, old and young
population shares). But the fiscal policy models also include non-discretionary grants as a
control variable. As mentioned in Section 3, these lump-sum grants are constitutionally
fixed and are thus not affected by the same incentives as discretionary grants. Finally, the
model for discretionary grants contains the index of mountainous areas which following
the remarks in Section 3 constitutes an important legal determinant of the size of
matching grants received. By interpreting the coefficient of this variable, please note
from Table 1 that this index takes on higher values for cantons with smaller mountain
areas and should thus be expected to have a negative sign on matching grants.
The analysis uses annual data for the 26 cantons from 1980 to 1998 inflated to the
year 1980. The empirical analysis is performed using a pooled cross-section time-series
model. We follow Feld and Kirchgässner (2001), who argue that despite the panel
structure of the data the inclusion of fixed effects in the cross-section domain is
inappropriate because the institutional variables vary only very little or remain even
constant over time in most cantons. This particularly holds for the federal index of
mountainous areas. Accordingly, cantonal intercepts do not make sense as the captured
impact on fiscal outcomes is either solely driven by the time variation or in case of time
invariant variables, fixed effects are likely to hide the effect of institutional variables and
render them insignificant. Moreover, year dummies are included, to circumvent time
dependency and the standard errors are corrected by a GMM method (Newey-West).
The nature of this empirical model suggests two additional steps that are required.
First, the second equation clearly determines matching grants which are an explanatory
variable in the first equation. It is this necessary to find instruments to cope with this
obvious endogeneity problem. The model as formulated above suggests to use the federal
index of mountainous areas as an instrument in the three fiscal policy equations. Second,
this model of four equations could either be estimated in a reduced form considering each
of the three fiscal policy equations as a separate decision. It could, however, also be
estimated as a structural model using systems estimation techniques. We follow the
second possibility and estimate a system of four equations by the GMM method.
The estimation results of the model are pretty straightforward. Matching grants per
capita are mainly influenced by income and the federal index of mountainous areas. Both
are significant at the 1 percent level and have a negative sign. The higher income, the
lower the amount of matching grants per capita a canton receives. The less mountain
areas a canton has, the lower are the matching grants it receives. Moreover, larger
cantons in population terms get significantly lower matching grants.
The matching grants a canton can expect to get in the future due to these
determinants significantly affect real cantonal spending and debt per capita, but do not
have any significant impact on the budget deficit per capita. As theoretically expected,
higher discretionary grants induce soft-budget constraints such that the cantons spend
more and are more heavily indebted. The estimation results of the control variables in the
three fiscal policy equations are pretty much in line with earlier results by Feld and
Kirchgässner (2001), Feld and Matsusaka (2003) and Schaltegger and Feld (2004). They
do not offer much of a surprise and are in line with theoretical hypotheses.
In addition to this simultaneous equations model, we have estimated further systems
in order to check the robustness of our results. In a first step, the model has been
extended by including political variables that capture interest group influence. Following
Feld and Schaltegger (2005) we include the share of the cantonal administrations’
employees in the total cantonal population as a proxy of the power of the cantonal
bureaucracy. Since the bureaucracy obtains a higher discretionary power by disposing of
additional public funds, it exerts a stronger demand for grants. Moreover, lobbying for
grants at the federal level requires the hiring of additional personnel in the state
bureaucracy. The respective increase in the number of bureaucrats further raises the
power of the bureaucracy. That’s why the bureaucracy is asking for an extension of the
federal grant system as well. We hence expect the state bureaucracy variable to exert a
positive impact on matching grants per capita. The larger the cantonal bureaucracy, the
more easily a canton can develop new spending projects specifically designed to capture
federal funds. The second proxy for the power of interest groups in the model is the share
of union members from total cantonal population. There is plenty of anecdotal evidence
in federal countries and also in Switzerland that states (cantons) facing structural shifts of
their economies attempt to shift the burden of structural economic reform to the whole
country. Third, the share of farmers from total employment is taken as a proxy for
interest group influence. Swiss farmers are known as one of the most influential interest
group in the country. It could thus well be that they successfully lobby for more grants
because this is indirectly helping them to keep their high incomes. However, farmers get
direct subsidies from the federal level such that they will not need to lobby for higher
grants. The impact of farmers is thus indeterminate. Political preferences of the cantonal
constituencies are controlled for by including an ideology proxy. Ideology is measured by
the relative strength of parties in the government. The stronger leftist parties, the higher
the value of this variable. In addition, we include a language variable indicating the share
of the German speaking population in a canton. While the interest group variables are
supposed to influence matching grants and may thus serve as additional instruments in
the fiscal policy equations, ideology and the regional dummy are also included as
explanatory variables in the fiscal policy equations. Estimating this extended model, the
impact of discretionary grants remains highly significant and positive in the spending
equation (t-statistics: 8.69) and in the debt equation (4.17). The estimation results can be
obtained from the authors upon request.
A second robustness checks consists in the inclusion of institutional variables in the
simultaneous equations model. According to the analyses on Swiss fiscal policy quoted
above, in particular fiscal referendums and formal fiscal constraints play a role for
cantonal fiscal policy. Including a dummy variable for fiscal referendums and an index
variable for formal fiscal constraints does not affect the main results on the impact of
discretionary grants on spending and borrowing. Again, spending and public debt are
significantly positively affected by expected (future) matching grants while discretionary
grants do not have any significant effect on cantonal deficits per capita.
Table 2: GMM system estimates for of discretionary grants and fiscal policy, 26 Swiss cantons,
1980-1998
Variables
Matching Grants
per Capita (t+2)
Lump-Sum Grants
per Capita
Income per Capita
Income per Capita
(t+2)
Population
Population (t+2)
Urban
Urban (t+2)
Old
Old (t+2)
Young
Young (t+2)
Unemployment
Rate
Unemployment
Rate (t+2)
Index of Mountainous Areas(t+2)
Observations
R2
Matching Grants
per Capita (t+2)
–
–
–
-21.959***
(2.70)
–
-0.216**
(2.05)
–
-142.006
(0.55)
–
-9.548
(0.38)
–
-32.962
(1.03)
–
53.525
(1.03)
-27.836***
(8.55)
442
0.473
Expenditure per
Capita
1.374***
(4.07)
-1.562
(1.10)
96.928***
(2.57)
–
Debt per Capita
Deficit per Capita
1.452***
(2.67)
-8.560***
(3.17)
138.214**
(2.03)
–
0.028
(0.47)
-0.607*
(1.68)
10.488*
(1.74)
–
1.223***
(3.64)
–
-3.003***
(5.33)
–
0.070
(1.09)
–
2512.219***
(3.40)
–
7200.474***
(6.53)
–
369.889***
(3.00)
–
116.097*
(1.85)
–
96.455
(1.00)
–
20.139**
(2.04)
–
-121.764*
(1.88)
–
-36.928
(0.32)
–
29.919***
(2.69)
–
105.748
(0.88)
–
284.719
(1.15)
–
91.564***
(3.84)
–
–
442
0.696
–
442
0.615
–
442
0.423
SER
461.98
905.71
1563.88
217.08
Note: : t-values are given in parantheses. The computed standard errors have been corrected for Newey
West’s heteroskedasticity and serial correlation consistent covariance matrix. All regressions contain 17
year-dummies whose coefficients are not reported. ***, **, * indicate significance at 1%, 5% and 10%
levels, respectively. R 2 is the adjusted coefficient of determination and SER is the standard error of
regression.
Table 3: GMM system estimates for of discretionary grants and fiscal policy, 26 Swiss cantons,
1980-1998
Variables
Matching Grants
per Capita (t+2)
Lump-Sum Grants
per Capita
Income per Capita
Income per Capita
(t+2)
Population
Population (t+2)
Urban
Urban (t+2)
Old
Old (t+2)
Young
Young (t+2)
Unemployment
Rate
Unemployment
Rate (t+2)
Index of Mountainous Areas(t+2)
Employees of the
cantonal administration (t+2)
Union members
(t+2)
Farmers (t+2)
Matching Grants
per Capita (t+2)
–
–
–
-31.354***
(4.68)
–
-0.176
(1.30)
–
41.930
(0.19)
–
-8.316
(0.41)
–
63.502
(1.55)
–
-69.982*
(1.95)
-16.959***
(5.77)
46865.87***
(5.34)
3690.921*
(1.84)
-10773.42***
(2.71)
Ideology
Ideology (t+2)
German Speaking
Cantons
German Speaking
Cantons (t+2)
Observations
R2
168.469***
(3.25)
-1.987
(1.35)
442
0.651
Expenditure per
Capita
1.955***
(8.69)
-3.567***
(2.76)
109.025***
(3.02)
–
Debt per Capita
2.112***
(4.17)
-8.674***
(3.28)
76.206
(1.14)
–
0.057
(1.33)
-0.850***
(3.26)
13.627**
(2.43)
–
-1.029***
(3.22)
–
-2.730***
(5.03)
–
0.100
(1.46)
–
2419.361***
(3.92)
–
7278.368***
(6.89)
–
282.216**
(2.13)
–
91.092*
(1.80)
–
10.560
(0.12)
–
24.076***
(2.88)
–
-147.329**
(-2.56)
–
-190.772
(1.58)
–
27.370**
(2.19)
–
-48.715
(0.40)
–
235.401
(0.94)
–
49.080**
(2.43)
–
Deficit per Capita
–
–
–
–
–
–
–
–
–
–
–
–
67.285
(0.48)
–
178.484
(0.75)
–
34.585
(1.26)
–
-5.160
(1.29)
–
6.385
(0.90)
–
-0.811
(1.19)
–
442
0.671
442
0.556
442
0.418
SER
375.88
942.06
1680.73
217.96
Note: : t-values are given in parantheses. The computed standard errors have been corrected for Newey
West’s heteroskedasticity and serial correlation consistent covariance matrix. All regressions contain 17
year-dummies whose coefficients are not reported. ***, **, * indicate significance at 1%, 5% and 10%
levels, respectively. R 2 is the adjusted coefficient of determination and SER is the standard error of
regression.
Table 4: GMM system estimates for of discretionary grants and fiscal policy, 26 Swiss cantons,
1980-1998
Variables
Matching Grants
per Capita (t+2)
Lump-Sum Grants
per Capita
Income per Capita
Income per Capita
(t+2)
Population
Population (t+2)
Urban
Urban (t+2)
Old
Old (t+2)
Young
Young (t+2)
Unemployment
Rate
Unemployment
Rate (t+2)
Index of Mountainous Areas(t+2)
Employees of the
cantonal administration (t+2)
Union members
(t+2)
Farmers (t+2)
Matching Grants
per Capita (t+2)
–
–
–
-29.393***
(4.18)
–
-0.200
(1.33)
–
2.077
(0.01)
–
-12.807
(0.53)
–
67.438*
(1.69)
–
-52.291
(1.41)
-15.555***
(5.45)
48890.96***
(5.53)
4123.318**
(2.32)
-10325.66**
(2.43)
Ideology
Ideology (t+2)
German Speaking
Cantons
German Speaking
Cantons (t+2)
Fiscal Referendum
143.769**
(2.53)
-2.193
(1.40)
Expenditure per
Capita
1.948***
(8.05)
-3.677***
(3.02)
113.507***
(3.22)
–
Debt per Capita
2.091***
(3.79)
-8.566***
(3.27)
77.027
(1.15)
–
0.029
(0.67)
-0.737***
(2.81)
11.584**
(2.11)
–
-0.591**
(2.00)
–
-2.121***
(3.74)
–
0.080
(1.16)
–
2170.125***
(4.19)
–
6890.359***
(6.81)
–
252.162*
(1.90)
–
227.824***
(4.42)
–
210.142**
(2.06)
–
16.185*
(1.84)
–
-41.607
(0.73)
–
-48.137
(0.39)
–
21.465*
(1.89)
–
-49.056
(0.40)
–
207.533
(0.82)
–
53.440***
(2.76)
–
Deficit per Capita
–
–
–
–
–
–
–
–
–
–
–
–
143.966
(1.15)
–
302.082
(1.29)
–
29.964
(1.05)
–
2.409
(0.53)
–
17.381**
(2.21)
–
-1.098
(1.56)
–
-952.842***
(4.96)
-1349.526***
(3.73)
66.812*
(1.66)
Fiscal Referendum
(t+2)
Formal Fiscal Restraints
Formal Fiscal Restraints (t+2)
Observations
R2
36.226
(0.33)
-20.440
(0.64)
442
0.646
–
–
124.468
(1.58)
–
154.727
(0.92)
–
442
0.705
442
0.574
–
-49.868***
(3.28)
–
442
0.431
SER
378.64
892.87
1646.22
215.65
Note: : t-values are given in parantheses. The computed standard errors have been corrected for Newey
West’s heteroskedasticity and serial correlation consistent covariance matrix. All regressions contain 17
year-dummies whose coefficients are not reported. ***, **, * indicate significance at 1%, 5% and 10%
levels, respectively. R 2 is the adjusted coefficient of determination and SER is the standard error of
regression.
References
BORGE, L-E. AND J. RATTSO. 1999. ASPENDING GROWTH WITH VERTICAL FISCAL IMBALANCE:
DECENTRALIZED GOVERNMENT SPENDING IN NORWAY 1880 - 1990.@ UNPUBLISHED.
FELD, L.P. and G. KIRCHGÄSSNER (2001), The Political Economy of Direct Legislation: Direct
Democracy in Local and Regional Decision-Making, Economic Policy 33 (2001), 329 –
367.
FELD, L.P. and J.G. MATSUSAKA (2003), Budget Referendums and Government Spending:
Evidence from Swiss Cantons, Journal of Public Economics 87 (2003), 2703 – 2724.
FELD, L.P. and C.A. SCHALTEGGER (2005), Voters as Hard Budget Constraints: On the
Determination of Intergovernmental Grants, forthcoming in: Public Choice 2005.
GARCIA-MILA, TERESA, TIMOTHY J. GOODSPEED, AND THERESE MCGUIRE. 2001. AFISCAL
DECENTRALIZATION POLICIES AND SUB-NATIONAL DEBT IN EVOLVING FEDERATIONS.@
UNPUBLISHED PAPER.
GOODSPEED, TIMOTHY J. “BAILOUTS IN A FEDERATION.” 2002. INTERNATIONAL TAX AND
PUBLIC FINANCE. AUGUST. 9: 409-421.
INMAN, ROBERT P. 2001. ATRANSFERS AND BAILOUTS: ENFORCING LOCAL FISCAL DISCIPLINE
WITH LESSONS FROM US FEDERALISM.@ IN RODDEN, J., G. ESKELUND, AND J. LITVACK,
EDS. FISCAL DECENTRALIZATION AND THE CHALLENGE OF HARD BUDGET CONSTRAINTS.
DRAFT.
JONES, M., P. SANGUINETTI, AND M. TOMMASSI. 1999. APOLITICS, INSTITUTIONS, AND PUBLICSECTOR SPENDING IN THE ARGENTINE PROVINCES.@ IN J. POTERBA AND J. VON HAGEN,
EDS. FISCAL INSTITUTIONS AND FISCAL PERFORMANCE. CHICAGO: UNIVERSITY OF
CHICAGO PRESS.
SCHALTEGGER, C.A. and L.P. FELD (2004), Do Large Cabinets Favor Large Governments?,
Evidence from Sub-federal Jurisdictions, mimeo, Philipps-Universität Marburg 2004.
VON HAGEN, JURGEN AND MATZ DAHLBERG. 2002. ASWEDISH LOCAL GOVERNMENT: IS THERE
A BAILOUT PROBLEM.@ UNPUBLISHED DRAFT.
WILDASIN, DAVID. 1997. AEXTERNALITIES AND BAILOUTS: HARD AND SOFT BUDGET
CONSTRAINTS IN INTERGOVERNMENTAL FISCAL RELATIONS.@ UNPUBLISHED.