Prof. Dr. Friedrich Schneider
Johannes Kepler Universität Linz
Institut für Volkswirtschaftslehre
Altenbergerstraße 69
A-4040 Linz-Auhof
Tel.: 0043-732-2468-8210, Fax: -8209
E-mail: [email protected]
http://www.econ.jku.at/schneider
7. Public expenditures in Canadian provinces:
An empirical study of politico-economic interactions
Kurs Public Choice
WS 2012/13
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
1 / 33
Public expenditures in Canadian provinces:
An empirical study of politico-economic interactions
CONTENT
1.
Introduction
2.
Review of Literature
3.
The Theoretical Model of Frey and Schneider
4.
Empirical Testing (Sensitivity Analysis)
5.
Summary and Concluding Remarks
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
2 / 33
1. Introduction
(1) Two widely accepted explanations of short-term
fluctuations of public spending suggest that public budgets
vary over time in a cyclical fashion.
i. One explanation, the electoral cycle, claims that public
expenditures will increase before an election and
decrease afterward.
ii. The other explanation, the partisan cycle, argues that
governments’ ideology shapes policies and outcomes.
(2) Consequently, alternation in power between various
political parties induces significant changes in the size of
public budgets: Left-wing parties favor public spending
increases while right-wing parties aim at budget reductions.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
3 / 33
2. Review of Literature
(1) The earliest empirical study of electoral cycles on budgetary outcomes is
usually attributed to Tufte (1978) for his work on US federal public
expenditures. Tufte‘s findings revealed, among other things, that
federal transfer payments to individuals were generally higher during
election years.
(2) In all cases (also in the studies of Nordhaus (1975) and MacRae (1977)),
it is assumed that elections create an incentive for governments to
behave opportunistically.
Because incumbent politicians wish to be re-elected, they will try to use
public policy instruments to improve economic conditions, thereby
boosting their popularity and their chances of victory in the upcoming
election.
Consequently, it is believed that public expenditures increase just before
an election, and decrease afterwards.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
4 / 33
2. Review of Literature (cont.)
(3) The partisan model developed by Hibbs (1977):
i. It postulates that ruling parties must implement policies
consistent with the interest of their core constituencies
that elected them. Consequently, politicians are driven
by ideological motives.
ii. It is commonly assumed that left-wing parties favor
public spending increases while right-wing parties aim at
budget reductions.
iii. Because the partisan composition of governments
changes over time, the size of public budgets should
fluctuate.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
5 / 33
3. The Theoretical Model of Frey and Schneider
Frey and Schneider offer a positive model of politico-economic interactions
between the economy and the policy:
(1) The model aims at describing the interdependency between the
government and the state of the economy.
(2) The government is one of the major actors of the model. Its objective is to
maximize its utility subject to a set of constraints. The government
derives utility by establishing public policies that meet its ideological
stance.
(3) However, the government can pursue its goals only if it can secure its
re-election. If the government fears defeat in the next general election, it
will undertake an expansionary budget policy to stimulate the state of
the economy and, therefore, its support among voters.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
6 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
(4) A second group of actors in the model are voters:
They, too, are assumed to maximize their utility subject to constraints.
Moreover, voters hold the government accountable for the state of the
economy and believe that it can use policy instruments to change it.
Consequently, they support the incumbent government if they are
satisfied with its economic performance.
(5) However, voters may find that gathering information about the true
competency of the government required time and money at costs that
outweigh the benefits of being well informed. For that reason, it
becomes rational for them to use broad macroeconomic indicators, such
as unemployment and inflation, to evaluate the performance of the
incumbent.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
7 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
(5) Whether a government can pursue its ideological goals or must address
the electoral constraint can be determined by using the concepts of
surplus and deficit of popularity.
(6) It is assumed that there is a critical level of popularity necessary to win
the next election. The critical level of popularity is a predefined value
that takes into account the characteristics of political institutions (e.g. the
number of competing parties) and the degree of risk aversion of the
governing party.
(7) If the actual level of popularity is greater than the critical value, the
government faces a surplus of popularity and is free to pursue its
ideological goals.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
8 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
(8) However, if the actual level of popularity is less than the
critical value, then the government needs to secure its reelection. The larger the deficit, and the closer the next
election, the more urgent is the need for the government to
act.
(9) Frey and Schneider pointed out that the interest of
bureaucrats lies primarily in the continuous expansion of
state activities, which is consistent with the observed
incremental approach of the fiscal decision-making
process.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
9 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
The model, formulated by Frey and Schneider, is:
ΔSpendingt = β0 + β1 ΔRevenuet-1 + β2 (1–Q)∙Electiont
+ β3 (1–Q)∙Popularity2t-1 + β4 Q∙Left∙Popularity2t-1
+ β5 Q∙Center∙Popularity2t-1
+ β6 Q∙Right∙Popularity2t-1 + εt
[Equation 1]
(1) The dependent variable (ΔSpending) is the annual
percentage change of total provincial public expenditures,
in constant Dollars.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
10 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
(2) The theoretically expected signs:
 A positive relationship between revenues and expenditures
is expected, thus β1 > 0.
 The variables Election and Popularity measure,
respectively, the amount of time that has elapsed since the
last general election, and the support received by the
government from the electorate, as expressed by opinion
polls.
Consequently, the estimated values of the coefficients are
expected to be positive, β2 > 0 and β3 > 0.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
11 / 33
3. The Theoretical Model of Frey and Schneider (cont.)
(3) It is generally assumed that left-wing governments favor
budget increases and right-wing governments seek to cut
public expenditures. Therefore, we expect β4 > 0, β6 < 0,
and β4 > β5 > β6.
(4) However, the assumption about the objective (to decrease
spending) of right-wing governments may be too restrictive.
A right-wing governments may not be in a position to cut
the total amount of public spending as much as they desire.
However, it should be expected that the values of the
estimated coefficients, associated with the measures of
ideology, respect the following conditions: β4 > 0, and
β4 > β5 > β6.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
12 / 33
4. Empirical Testing
Table 1: Descriptive statistics of variables used in the regressions (N=65).
Variable
Mean
Standard deviation
Minimum value
Maximum value
D Spending
1.966
4.784
-13.79
14.83
D Revenue
2.695
6.198
-24.16
19.02
Q
0.477
0.503
0.00
1.00
Popularity
0.228
22.746
-44.20
50.80
Election
32.692
16.654
6.00
60.00
Left
0.262
0.443
0.00
1.00
Center
0.169
0.378
0.00
1.00
Right
0.569
0.499
0.00
1.00
Budgetary Balance
-7.869
7.627
-26.83
13.82
D Transfers
1.571
8.730
-13.27
18.52
D Personal Income
1.677
3.081
-3.98
8.60
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
13 / 33
4. Empirical Testing (cont.)
Table 2:
Regression estimates for the Frey
and Schneider hypothesis - Part 1
WS 2012/13
Variable
Equation 1
D Revenue
0.2214 (2.68)***
(1–Q)∙Election
0.1113 (4.70)***
(1–Q)∙Popularity2
-0.0009 (-0.80)
Q∙Left∙Popularity2
0.0122 (2.75)***
Q∙Center∙Popularity2
0.0055 (3.63)***
Q∙Right∙Popularity2
0.0040 (4.06)***
Election
-
Left
-
Center
-
Right
-
Popularity2
-
Left ∙Popularity2
-
Center∙Popularity2
-
Right∙Popularity2
-
Budgetary Balance2
-
D Transfers
-
D Personal income2
-
Constant
-1.8108 (-2.37)**
N
65
Buse R2
0.4005
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
14 / 33
4. Empirical Testing (cont.)
(1) As indicated by the coefficient of determination, the model
explains 40% of the variance of annual variation of
provincial public spending.
(2) Furthermore, five of the six estimated parameters have the
expected sign and are statistically significant at a level
of 1%.
(3) A one percentage-point increase in real revenue increases
public provincial budget by less than one quarter of a point
(β1 = 0.2214). Therefore, the impact of the budgetary
constraint seems of little magnitude, at least if balanced
budgets are the target.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
15 / 33
4. Empirical Testing (cont.)
(4) The estimates also confirm the presence of a positive
relationship between public spending and the proximity of
the next election when the government faces a deficit of
popularity. The closer the next election, the more the
government will increase public expenditures.
(5) If a government is struggling with a deficit of popularity
during a whole year,
 it will increase its budget on average by about 1.3
percentage points during its first year in office
(β2 = 0.1113 × 12 months), and
 up to 6.7 percentage points during its last year (60
months).
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
16 / 33
4. Empirical Testing (cont.)
(6) The findings reveal that left-wing governments increase
public expenditures more than twice as much as centrist
and right-wing governments do (β4 > β5 > β6) when they have
a surplus of popularity. Increases in public expenditures
are also higher under centrist governments compared to the
right (β5 > β6).
(7) The estimates indicate, for instance, that a left-wing
government, having a 10 percentage-point lead in opinion
polls, will increase public expenditures by 1.22% per year
compared to 0.55% for a centrist government and 0.40%
for a right-wing government.
WS2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
17 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(8) Removing all indicators of popularity (Popularity and Q)
from equation [1], one can estimate the following equation
and verify if the effect of the two cycles are independent of
each other.
If they are, equation [2] should produce estimates that have
the predicted sign, are statistically significant and explain a
larger amount of the variance in public expenditures.
ΔSpendingt =
β1 ΔRevenuet-1 + β2 Electiont + β3 Leftt-1
+ β4 Centert-1 + β5 Rightt-1 + εt
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
[Equation 2]
18 / 33
4. Empirical Testing (cont.)
– Sensitivity Analysis
Table 2:
Regression estimates for the Frey
and Schneider hypothesis – Part 2
WS 2012/13
Variable
Equation 2
D Revenue
0.2022 (1.89)*
(1–Q)∙Election
-
(1–Q)∙Popularity2
-
Q∙Left∙Popularity2
-
Q∙Center∙Popularity2
-
Q∙Right∙Popularity2
-
Election
0.0551 (1.62)
Left
-1.7105 (-1.02)
Center
0.9933 (0.85)
Right
-0.1705 (-0.13)
Popularity2
-
Left ∙Popularity2
-
Center∙Popularity2
-
Right∙Popularity2
-
Budgetary Balance2
-
D Transfers
-
D Personal income2
-
Constant
-
N
65
Buse R2
0.1554
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
19 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(9) However, we find that the explanatory power of the
regression drops to about 16% (compared to 40% for
Equation [1]) and that none of the estimated coefficients
are statistically significant at the 5% level.
(10) Moreover, some estimated coefficients linked with ideology
display values contrary to prediction. For instance, our
findings indicate that left-wing governments reduce public
spending ten times more than right-wing governments.
Clearly, equation [2] does not provide support for a
refutation of the Frey and Schneider hypothesis.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
20 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(11) Our initial findings could also suggest that the behavior of governments
might be influenced by the level of popularity, but not by the presence
of a surplus or a deficit of popularity.
The underlying assumption would be that the electoral and the
partisan cycles are independent from each other.
Dropping only the variable Q from equation [1] and assuming that
there is no interaction between the level of popularity of the
government and its ideology, the model can be rewritten in the
following way:
ΔSpendingt = β1 ΔRevenuet-1 + β2 Electiont + β3 Leftt-1
+ β4 Centert-1 + β5 Rightt-1 + β6 Popularity2t-1 + εt
[Equation 3]
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
21 / 33
4. Empirical Testing (cont.)
– Sensitivity Analysis
Table 2:
Regression estimates for the Frey
and Schneider hypothesis – Part 3
WS 2012/13
Variable
Equation 3
D Revenue
0.2019 (1.99)*
(1–Q)∙Election
-
(1–Q)∙Popularity2
-
Q∙Left∙Popularity2
-
Q∙Center∙Popularity2
-
Q∙Right∙Popularity2
-
Election
0.0651 (1.92)*
Left
-1.6413 (-0.98)
Center
0.2268 (0.18)
Right
-0.5585 (-0.40)
Popularity2
0.0011 (1.30)
Left ∙Popularity2
-
Center∙Popularity2
-
Right∙Popularity2
-
Budgetary Balance2
-
D Transfers
-
D Personal income2
-
Constant
-
N
65
Buse R2
0.1815
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
22 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(12) Devine (1985) has suggested that popularity can nonetheless
interact with ideology. If this is the case, the model of
estimation can be represented by equation [4]:
ΔSpendingt = β0 + β1 ΔRevenuet-1 + β2 Electiont
+ β3 Left∙Popularity2t-1
+ β4 Center∙Popularity2t-1
+ β5 Right∙Popularity2t-1 + εt
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
[Equation 4]
23 / 33
4. Empirical Testing (cont.)
– Sensitivity Analysis
Table 2:
Regression estimates for the Frey
and Schneider hypothesis – Part 4
WS 2012/13
Variable
Equation 4
D Revenue
0.1980 (1.90)*
(1–Q)∙Election
-
(1–Q)∙Popularity2
-
Q∙Left∙Popularity2
-
Q∙Center∙Popularity2
-
Q∙Right∙Popularity2
-
Election
0.0619 (1.77)*
Left
-
Center
-
Right
-
Popularity2
-
Left ∙Popularity2
0.0010 (1.10)
Center∙Popularity2
0.0024 (1.53)
Right∙Popularity2
0.0014 (1.18)
Budgetary Balance2
-
D Transfers
-
D Personal income2
-
Constant
-0.7147 (-0.56)
N
65
Buse R2
0.1676
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
24 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(13) The estimates for equations [3] and [4] are presented in
parts 3 and 4 of Table 2. These additional estimates do not
provide better explanations for changes in public spending
than equation [1].
(14) In both cases, the explanatory power is lower, and several
estimated coefficients have a sign or value contrary to
predictions, and are not statistically significant.
(15) Overall, the sensitivity analysis presented here gives
additional support to the Frey and Schneider hypothesis:
The existence of an electoral or a partisan cycle is conditional
to the presence of a surplus or a deficit of popularity.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
25 / 33
4. Empirical Testing (cont.) – Sensitivity Analysis
(16) Finally, since nearly 50% of total provincial revenue comes
from personal income and consumption taxes (in addition:
about 15% is financed by federal transfers, 10% by
investment income, and 6% by corporate taxes),
it seems that the capacity of taxpayers to finance public
budgets
should
also
be
taken
into
account.
For this reason, three new indicators of the budgetary
constraint can be substituted for the variable ΔRevenue.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
26 / 33
4. Empirical Testing (cont.) – Sensitively Analysis
(17) The model can now be estimated as follows:
[Equation 5]
ΔSpendingt = β0 + β1 (1–Q)∙Electiont + β2 Q∙Left∙Popularity2t-1
+ β3 Q∙Center∙ Popularity2t-1 + β4 Q∙Right∙Popularity2t-1
+ β5 Budgetary Balance2t-1 + β6 ΔTransferst + β7 ΔPersonal Income2t-1 + εt
where
 Budgetary Balance is the ratio of public provincial deficit in
percentage of total public expenditures;
 Transfers is the annual variation in percentage of federal transfers (in
real terms) to the provincial government; and
 ΔPersonal Income is the annual variation in percentage of provincial
personal income (also in real terms).
It is expected that all coefficients related to the budgetary constraint will
be positive (β5 > 0, β6 > 0, β7 > 0). Results are shown in table 2 part 5.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
27 / 33
4. Empirical Testing (cont.)
– Sensitivity Analysis
Variable
Equation 5
D Revenue
-
(1–Q)∙Election
0.0872 (4.86)***
(1–Q)∙Popularity2
Table 2:
Regression estimates for the Frey
and Schneider hypothesis – Part 5
WS 2012/13
Q∙Left∙Popularity2
0.0102 (2.97)***
Q∙Center∙Popularity2
0.0032 (2.41)**
Q∙Right∙Popularity2
0.0027 (3.91)***
Election
-
Left
-
Center
-
Right
-
Popularity2
-
Left ∙Popularity2
-
Center∙Popularity2
-
Right∙Popularity2
-
Budgetary Balance2
0.0128 (4.49)***
D Transfers
0.1771 (3.46)***
D Personal income2
0.0871 (3.53)***
Constant
-0.2503 (-0.37)
N
65
Buse R2
0.6208
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
28 / 33
4. Empirical Testing – Cont. – Sensitively Analysis
(18) The modified model now explains about 60% of the
variance in annual variations of provincial public spending.
(19) In addition, all of the seven estimated parameters have the
expected sign and are statistically significant at the 1% level
for all parameters, except
Q · Center · Popularity²
(significant at the 5% level).
(20) All of the fundamental assumptions identified in the initial
model (equation [1]) are supported when the three new
variables, related to the budgetary constraint, are
incorporated in the model (equation [5]).
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
29 / 33
Table 3: Additional Results: Disaggregating Provincial Spending
Variable
Health
Education
Social
Services
Resources &
Industry
Transport &
Communication
General Government
services
(1–Q)∙Election
0.0422
(2.42)**
0.0369
(0.85)
0.0258
(0.79)
0.2742
(1.80)*
0.0866
(1.19)
-0.0336
(-0.55)
Q∙Left∙Popularity2
0.0082
(2.70)***
0.0108
(0.81)
0.0138
(2.39)**
-0.0058
(-0.28)
0.0139
(1.69)*
0.0080
(0.51)
Q∙Center∙Popularity2
0.0014
(0.86)
-0.0029
(-0.82)
0.0020
(0.80)
0.0104
(1.10)
0.0068
(1.38)
0.0001
(0.01)
Q∙Right∙Popularity2
0.0013
(1.51)
0.0013
(0.57)
0.0028
(1.37)
0.0028
(0.62)
0.0013
(0.51)
0.0036
(1.06)
Budgetary Balance2
0.0113
(4.44)***
0.0101
(1.60)
0.0091
(1.79)*
0.0521
(2.39)**
0.0221
(2.25)**
-0.0099
(-1.07)
D Transfers
0.13787
(3.49)***
0.0993
(0.88)
0.1408
(1.79)*
0.8282
(1.92)*
0.0036
(0.01)
0.1712
(1.05)
D Personal income2
0.0806
(3.40)***
0.2248
(3.52)***
0.0223
(0.45)
-0.0419
(-0.24)
0.0503
(0.36)
0.1232
(1.49)
Constant
0.96220
(1.42)
-0.4094
(–0.31)
3.6515
(2.61)**
0.6327
(0.09)
-1.7268
(-0.64)
-2.0444
(-1.06)
N
65
65
65
65
65
65
Buse R2
0.5143
0.3068
0.1924
0.1924
0.1210
0.0944
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
30 / 33
Table 4: Additional Results: Disaggregating Provincial Spending
Variable
Protection
Environment
Recreation &
Culture
Regional
Planning
Labor &
Immigration
Total spending net
of debt service
(1–Q)∙Election
0.0390
(0.87)
0.1444
(0.92)
0.1029
(1.08)
0.1687
(1.09)
0.2587
(1.60)
0.0928
(4.20)***
Q∙Left∙Popularity2
0.0056
(0.56)
-0.0090
(-0.41)
-0.0060
(-0.43)
0.0199
(0.62)
0.0718
(2.36)**
0.0107
(2.62)**
Q∙Center∙Popularity2
0.0051
(1.22)
0.0016
(0.14)
0.0056
(0.76)
-0.0144
(–1.19)
0.0355
(2.41)**
0.0037
(2.27)*
Q∙Right∙Popularity2
0.0028
(1.24)
-0.0028
(-0.42)
-0.0021
(-0.62)
-0.0068
(-1.18)
0.0103
(1.61)
0.0024
(2.83)***
Budgetary Balance2
0.0122
(1.91)*
0.0561
(2.15)**
0.0391
(3.07)***
0.0451
(2.26)**
0.0534
(2.53)**
0.0180
(5.19)***
D Transfers
-0.0022
(-0.02)
0.5095
(1.29)
0.4560
(1.66)*
0.1475
(0.34)
-0.0074
(-0.02)
0.2264
(3.84)***
D Personal income2
0.1363
(2.18)**
-0.0680
(-0.33)
0.0231
(0.14)
0.2670
(0.93)
-0.2765
(-1.01)
0.0984
(3.33)***
Constant
-0.1055
(-0.07)
10.557
(1.59)
2.7042
(0.71)
-0.4614
(-0.09)
1.6467
(0.31)
-0.4710
(-0.54)
N
65
65
65
65
65
65
Buse R2
0.2662
0.1281
0.1678
0.1428
0.1887
0.6155
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
31 / 33
5. Summary and Concluding Remarks
(1) The purpose of this paper was to investigate whether the hypothesis,
formulated by Frey and Schneider, on the interaction between the
electoral cycle and the partisan cycle could be validated empirically for
Canadian provinces.
(2) Using Canadian provincial public spending over the period 1983-1995,
we found empirical support for the Frey and Schneider explanation,
validated by the sensitivity analyses.
(3) This study is the first attempt to test the Frey and Schneider hypothesis at
the subnational level.
(4) Does Canada constitute a unique case? At first sight, there seems to be
no reason why this should be the case. Under democratic rules, it can be
expected that politicians are driven by electoral and partisan motives.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
32 / 33
Source
Tellier, Genevieve,
Public Expenditures in Canadian Provinces:
An Empirical Study,
Public Choice 126/3, 2006, pp.367-385.
Franzese, R.J.J. (2002),
Electoral and Partisan Cycles in Economic Policies
and Outcomes,
Annual Review of Political Science. 5: 369-421.
WS 2012/13
© Prof. Dr. Friedrich Schneider, University of Linz, Austria
33 / 33