7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 1/47
Political effects on the allocation of public expenditures:
Empirical evidence from OECD countries
Niklas Potrafke1 Humboldt University Berlin
This Version: April 20, 2008
Abstract
This paper examines the effects of political determinants on the allocation of public expenditures.
We analyse two data sets of different expenditure categories (COFOG) covering a broad time
period: An OECD panel from 1970 to 1989 as well as from 1990 to 2005. We find that national
policy had stronger impacts till the beginning of the nineties. Leftist governments set other
priorities than rightwing governments, but this required that they had a majority in parliament.
Electoral effects are detected as endogenous and the size of coalition emerges as unimportant.
Keywords: public expenditures, budget composition, partisan politics, political budget cycle,
panel data
JEL Classification: D72, H50, H61, C23
Acknowledgements: We thank Roland Vaubel, Ulrich Oberndorfer, Charles B. Blankart,
Georgios Chortareas, Frank Somogyi, Gunther Markwardt and the participants of the WPCS2007, of the CESifo Workshop “Macroeconomics” 2007 and of the DIW Workshop“Macroeconometrics” 2007 for helpful comments, hints and suggestions. All errors are our own.
1 Humboldt University Berlin, Department of Economics and Management Science, Institute of Public Finance,
Competition Policy and Institutions, Spandauer Strasse 1, D-10178 Berlin, Germany, Phone: + 49 30 2093 5788,Fax: + 49 30 2093 5697. Email: [email protected]
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 2/47
Niklas Potrafke
2
1. Introduction
This paper examines the political determinants of the budget composition in OECD countries.
The essence is that the allocation of expenditures is an interesting object of investigation in
national fiscal policy because it covers the politicians’ real room of manoeuvre. Irrespective of the
amount of revenues and expenditures and the need to take care of given allegiances, each
government has to choose the purposes for which it will spend the revenues. It will prioritise.
However, a respective empirical analysis requires data that mirror different policy fields by single
expenditure categories. The more comprehensive the classifications of government expenditure,
the higher the probability that single policy effects are undetected. They might simply
compensate each other. Therefore this paper refers to the classification of functions of
government (COFOG). We analyse the data set by Sanz and Velázquez (2007) and Gemmell et
al. (2007) covering the period from 1970 to 1989 as well as the recent data set from the OECD
for the period from 1990 to 2005 and thereby start examining the budget composition of general
governments more comprehensively than previous literature does.
Analyzing the budget composition enjoys a remarkable popularity in the very recent empirical
literature. Sanz and Velázquez (2007) investigate the role of aging on the allocation process and
detect that aging is the main driving force of the growth of government spending in OECD
countries. Shelton (2007) provides an extensive analysis of different impact factors like
population, country size, fragmentation, income, income inequality, political rights and
institutions. He distinguishes between different levels of government (central, and local) and
considers more than 100 countries. Dreher et al. (2007) analyze whether globalization has
affected the composition of public expenditures also using the KOF index of globalization
(Dreher (2006a) and Dreher et al. (2008)) and do not find any influence. Gemmell et al. (2007)
detect that foreign direct investment significantly shifted the expenditure composition towards
social spending, whereas trade did not. Hence it is still not clear whether budgets are influenced
by international factors. But they may be driven by national policy and political determinants like
the different attitudes and number of parties in the government, the timing of elections etc.
There are a few further studies testing for political effects on the allocation of public
expenditures, but first, they do not examine the detailed budget composition (COFOG) as the
studies listed above. Bräuninger (2005) develops a partisan model of government expenditure and
also provides empirical tests in an OECD framework in the period from 1971 to 1999. He finds
that the actual spending preferences of parties matter whereas they do not indicate that parties of
the left consistently differ from parties of the right in their spending behaviour. Tsebelis and
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 3/47
Niklas Potrafke
3
Chang (2004) analyse the impact of veto players on the budget composition. Their results show
that it changes at a slower pace by multiparty governments. However, they choose a different
model set up taking the change in the structure of budgets as dependent variable (budget
distance). Furthermore, there are some single country studies examining the budget composition
in Germany in more detail: Bawn (1999), Galli and Rossi (2002), König and Tröger (2005) and
Potrafke (2006). In sum, they find that the budget composition is driven by political
determinants. Regarding the political variables the current paper extends the previous research by
first examining the interaction of government ideology and the fact if the respective governments
have a majority in parliament. We expect somewhat mitigated ideology effects under minority
governments. Second we apply an extended variable set of predetermined and endogenous
elections years following Shi and Svensson (2006).
The remainder of the paper is organised as follows: Section 2 provides the institutional
background originating from the theory of political economics. Section 3 presents the data and
discusses their time series properties. In section 4 the empirical model is set up and the political
variables are described. Section 5 discusses the estimation results and section 6 concludes the
analysis.
2. Institutional background2.1 Political business cycles, partisan approach and government types
The issue of the paper is to test for the effects of election years, the ideological party composition
as well as the type of government on the allocation of public expenditures. The impacts of these
variables on economic activity stem from a huge and model based literature of political
economics. In this paper, our emphasis is not to find evidence for a single theoretical model.
Instead we will very briefly repeat the main ideas of respective (well known) theoretical work
establishing a basis for the following empirical analysis.
First, the political business cycle approaches and the partisan theory clarify how politicians try to
influence economic outcome. One implication of the theories by Nordhaus’ (1975) and Rogoff
and Sibert (1988) among others is that all politicians will implement the same policy. Ideology
does not matter. Policies will converge. In addition, they imply a particular pattern between
elections on the one hand and the impacts of economic policy on the other hand. Nordhaus
(1975)’ opportunistic school asserts that politicians fool the public just to win elections. They will
boost economic activity right before elections. The rational political business cycle theory by
Rogoff and Sibert (1988) among others criticizes this modelling for using adaptive expectations
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 4/47
Niklas Potrafke
4
and introduces rational expectations instead. In this approach, information asymmetries play a
role as a source of the electoral cycles. The incumbent tries to exploit his information advantage
by signalling his economic competence before elections. This scientific debate is still alive.
Recently, Shi and Svensson (2006) show that politicians may behave opportunistically even if the
voters know all the government programs, but some individuals are uninformed. Alt and Lassen
(2006) point out that the higher the transparency of the political process, the lower the
probability to behave opportunistically. Finally, we conclude from these approaches that election
years will affect the budget composition so that the preferences of the median voter are fulfilled.
In contrast, the partisan approach focuses on the strong impact of party ideology. As a result,
platforms and policies do not converge. Instead, leftwing and rightwing politicians will provide
different policies by concentrating on the preferences of their partisans. The leftist party appeals
more to the labour base and promotes expansionary policies, whereas the rightwing party appeals
more to capital owners, and is therefore more concerned with reducing inflation. This holds for
both sub-approaches of the partisan theory - for the classical approach of Hibbs (1977), and for
the rational approach of Alesina (1987). The literature on the partisan theory has become more
abundant during the last decades (e.g. Alesina et al. 1997). In fact, Belke made several important
contributions to this literature and introduced the hysteris-augmented rational partisan approach
(Belke 1996, 1997a, 1997b, 1997c and 2000). Overall, our hypothesis is that party constellationand respective ideologies of the governments affect the budget composition.2 3
Another political determinant stems from the literature on fiscal policy. It arises from the
“common pool problem” discussed, e.g., by Weingast et al. (1981) and implies that decision costs
increase with the number of decision makers. This also refers to the logic of logrolling. For
example, the collective action literature implies that the more dispersed the decision-making
authority, the higher the budget deficit. Also the amount of government expenditures is expected
to be higher the more parties form a government. Tsebelis (1995) is associated with this so called
early veto player theory. He claims that the potential of a policy change decreases with the
number of veto players, the lack of congruence (dissimilarity of policy positions among veto
players) and the internal cohesion (similarity of policy positions among the constituent units of
2
This implies that there is divergence of policies and platforms. Theoretically, in a simple two party model, ideologymust over compensate the vote maximizing effect in this case. In a multi party model, manifoldness and traditions ofthe parties are assumed to avert policy convergence. See, e.g., Mueller (2003): Chapters 11-13 and Persson and
Tabellini (2000): Chapters 3 and 5 for an overview of the respective fundamental literature on party competition. Thecurrent paper is not the right place to discuss the impact of ideology, what it means or where and it comes from.3
The literature on the partisan theory has become much more comprehensive during the last decades, of course (see,e.g., Alesina et al. (1997)). We just highlighted the beginnings but still highly relevant papers of the debate.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 5/47
Niklas Potrafke
5
each veto player) of these players. However, this is just the one side of the coin. The second
effect claimed by most recent applications points out that policy stability increases with the
number of veto players (Tsebelis (2002)). There might be a lock-in effect: Countries with more
veto players have consistently low or, consistently high deficits. Stability might increase with the
number of coalition partners. Overall, the final impact remains as an empirical question. As
coalition partners also have to find agreements how they will spend their revenues, we expect that
the type of government, namely the number of coalition partners as well as the fact if the ruling
government has a majority in parliament (minority government) affect the budget composition.
Beyond this, we will not discuss other impacts and interactions4 any further. For example,
national political influences on economic issues might be mitigated due to globalization etc.
Garrett (1998: Chapter 2) discusses in detail how domestic policy might work in the globaleconomy. Potrafke (2007a) explicitly examines the interaction of government ideology and
globalization and tests its impacts on social expenditures. However, the current paper focuses on
national policy. Another caveat against the application of the respective theory to the budget
structure might be that newly-elected governments are not able to change the current budget.
Tsebelis and Chang (2004: 457 f.) provide a convincing discussion, why the current government
is responsible for the realization of the budget because it has the means to alter the existing
budget: “First, finance ministers in most EU countries (the exceptions being Finland, the
Netherlands, Spain and Sweden) can either block expenditure or impose cash limits. They also
have the power to allow funds to be transferred between chapters, and the disbursement of the
budget in the implementation stage is subject to finance ministers’ approval. Second, there is a set
of formal rules that enables governments to deal with unexpected expenditure and revenue
shocks…”
2.2 The political economy of the budget composition
This subsection supports the application of the theories listed above on the budget composition
by presenting already existing theoretical models as well as empirical evidence on particular
expenditure categories. Bräuninger (2005) offers a partisan model of government expenditure. He
assumes that for electoral reasons partisan actors generally prefer more spending to less but also
experience electoral costs from higher experience due to the increasing tax burden. Further, he
wants to emphasize the distinction between the ideological identity of governmental actors and
4
Interactions between the ideological party composition of the government and the number of coalition partners
could be considered in more detail. But then, the judging of respective coalition types becomes much morecomplicated and further assumptions have to be made. Tsebelis and Chang (2004) create multidimensional indices intheir veto player model. These indices also take into account the ideological distances between the parties in eachgovernment.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 6/47
Niklas Potrafke
6
their programmatic policy and spending preferences. Two approaches are considered: A median
legislator model and a veto player model. Basically, the model predicts that the higher spending
preferences of the political actors for a certain issue, the higher will indeed be the real spending.
Drazen and Eslava’s (2005) model of the political budget cycle focuses on the electoral effects.
Technically, the idea of targeted expenditures is close to earlier work from Lindbeck and Weibull
(1987) and Dixit and Londregan (1996). First, voters have different preferences regarding the
types of spending. The same holds for the politicians. But the latter might shift the composition
of the spending towards the goods voters prefer. Thereby, the politicians could signal that their
preferences are close to those of the voters, implying they will choose high-post election
spending on the same goods. In equilibrium, “a political budget cycle will exist, in which: 1)
expenditures targeted to voters are expected to be higher in an election than a non-election
period; and 2) swing voters will rationally vote for an incumbent who provides higher targeted
expenditures even though they know that such expenditures may be electorally motivated”
(Drazen and Eslava (2005: 14)). Considering the framework of asymmetric information about the
electoral environment, the model predicts even more room for the politicians to influence the
outcome of elections by providing more targeted expenditures prior to elections.
All-embracing hypotheses regarding the detailed way of allocating the expenditures by different
parties are not easy to formulate. It is simply impossible to classify all the COFOG expenditurecategories regarding these variables. Also Drazen and Eslava (2005: 3) state: “Obviously a
classification of government expenditure into targeted and non-targeted expenditures is not
readily available or straightforward”. However, more concrete hypotheses might be necessary
because of fundamental reasons in empirical work. Further, referring to Bräuninger (2005) and
Drazen and Eslava (2005) the categories might be grouped regarding items of leftist and
rightwing governments as well as targeted to the median voter. But the current paper will test for
ideological (leftist versus rightwing) and election year effects (affecting the median voter)
together. Hence a grouping regarding both effects will be difficult. In addition, some types of
expenditure are more (in)elastic than others and may be subject to long run contracts, e.g.,
defence.
There is further literature on the impacts of political determinants on single expenditure
categories. Boix (1997) claims that leftist governments are expected to spend more on education
than conservative governments. He lastly concludes from his empirical analysis from 1970 to
1989 that “a great deal of public spending on education is driven by “demand” factors – mainly
the demographic structure of the country” (Boix (1997: 835)). There were partisan effects
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 7/47
Niklas Potrafke
7
regarding public spending on education in the sixties and eighties but not in the seventies. Hicks
and Swank (1992) provide theory and evidence how policy influences welfare spending in
industrialized countries. For example, following the social democratic corporatist perspective,
leftwing and center (non-rightwing) party leadership of government generates higher welfare
effort than rightwing and “intermediate” party leadership. Leftist governments might even be
pressured to moderate prowelfare policies by the opposition. They further explore the hypothesis
that preelection welfare effort tends to exceed postelection effort (Hicks and Swank (1992: 659
f.)). Further empirical evidence detects higher social expenditures under leftist than rightwing
governments till the end of the “cold war” in 1990, but more responsibility in the 90ies (e.g.,
Kittel and Obinger 2003, Potrafke 2007a, Dreher 2006b does not find significant effects in the
period from 1970 to 2000). Nincic and Cusack (1979) examine the political economy of US
military spending and find an electoral cycle with higher military expenditures before elections.
Correa and Kim (1992) provide an overview of the literature on defence expenditure in the USA
and the USSR and conclude referring to the literature as well as their own empirical research that
defence expenditures in the USA are driven by political variables. They even find higher defence
expenditures under democratic presidents but admit that this finding contradicts the position
usually attributed and “could change if a longer time period were included in the analysis”
(Correa and Kim (1992: 168)). Leftist and rightwing governments might also have different
spending preferences regarding health. Immergut (1992: 1) states: „National health insurancesymbolizes the great divide between liberalism and socialism, between the free market and the
planned economy…Political parties look to national health insurance programs as a vivid
expression of their distinctive ideological profiles and as an effective means of getting
votes…National health insurance, in sum, is a highly politicized issue.” Hence we further expect
higher public health expenditures under leftist than rightwing governments. Potrafke (2007b)
analyses an OECD panel and confirms this claim for the period from 1970 to 1990. However,
things change in the period from 1991 to 2004.
We will formulate hypotheses for these (core-) categories for which mappings with respect to the
single theories seem to be clear-cut and leave open the others. Thus Table 1 already presents the
different categories of public expenditure (COFOG) and easily acts as link to the next section.
The signs “+” and “–“ indicate an expected increasing or decreasing effect of the political
variables on the categories, respectively.
Table 1 about here
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 8/47
Niklas Potrafke
8
3. Data
3.1 Two data sets
There is no unique data set available yet, that classifies public expenditures of the general
government by so called COFOG (Classification of the Functions of Government) functions and
types for a longer time period, e.g., from 1970 till present. Therefore, we examine two different
data sets classifying public expenditures of the general government5: The one composed and used
by Sanz and Velázquez (2007) and Gemmell et al. (2007) covering the period from 1970 to 1989
as well as the recent data sets from the OECD for the period from 1990 to 2005. The first one
contains yearly data for 23 OECD countries: Australia, Austria, Belgium, Canada, Denmark,
Finland, France, Germany 6, Greece, Iceland, Ireland, Italy, Luxembourg, the Netherlands, New
Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the UK and the USA (balanced panel).
The second data set from the OECD contains yearly data for the total expenditure structure of
13 OECD countries from 1990 to 2005. The panel is unbalanced. There are yearly data available
for Belgium, Denmark, Finland, Ireland, Luxembourg, the UK as well as the USA for the period
from 1990 to 2005. Data for Germany are available from 1991 to 2005, for Italy from 1990 to
2004. Austria, France and Sweden can be included with data running from 1995 to 2005. Lastly,
there are data for Japan from 1996 to 2005.7 Further, we will not consider any country with fewer
observations, so that at least some variation in time elapsed is mirrored and policy changes might
occur.
The examined data are public expenditures classified by so called COFOG in both cases, but
however, they differ in some respects. First, Sanz and Velázquez (2007) combined the two
categories of the original classification “General public services” and “Public safety and order” to
one category named “Public services”. We proceed in the same way for the second data set from
1990 to 2005 to make the results somewhat comparable. Hence this expenditure category refers
to the provision of publicly provided goods. Second, the new OECD classification includes a
category called “Environmental protection”.8 This differs from previous classifications so that
this category is not included in the data set by Sanz and Velázquez (2007). Instead Sanz and
Velázquez (2007) split “Transport and communications” from “Economic Services” because
“Transport and communications” are a kind of investment. Hence overall, both data sets
5
The data refer to the general government. Hence, we do not distinguish between the different jurisdictions in thesingle countries and take the institutional background into account as Shelton (2007) and Potrafke (2006) do. 6Germany will be excluded in our basic regressions because of missing data on some control variables. Section 5.3
will point out that our inferences are not sensitive to this exclusion.7 There are also data for South Korea from 1996 to 2005. However, South Korea is a presidential system, so that thecurrent analysis of the political variables is not applicable.8
Most of the expenditures for “Environmental protection” were classified via “Housing” due to the formercategorization.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 9/47
Niklas Potrafke
9
distinguish between nine different expenditure categories where eight of them are named the
same in both data sets. Lastly, both data sets exclude interest payments. We will use the different
expenditure categories as dependent variables for the examination of the allocation of
expenditures across the countries. The appendix contains a detailed description of the single
expenditure categories due to the classification system and descriptive statistics9 of all variables.
3.2 Time series properties
The time series properties of the single variables emerge as very crucial for the econometric
specifications. The inclusion of non-stationary variables in the model might cause spurious
regression. Testing for stationarity of the time series, we apply a battery of panel unit root tests.
The advantage of the panel unit root tests compared to the univariate counterparts is to gain
statistical power. However, the tests to a panel also relate to asymptotic theory and therefore lose
power in small samples. Breitung and Pesaran’s (2005) overview on unit roots and cointegration
in panels points out that the respective tests refer to samples where the time dimension (T) and
the cross section dimension (N) are relatively large. However, we will carefully apply the battery
of respective tests, but will also stress that the results should be handled somewhat carefully due
to the small sample sizes. The appendix provides our several test results in detail and comments
on the chosen procedures. In conclusion, we will estimate the model in first differences.
4. The empirical model
As all categories described sum up to total expenditure and the government has to choose for
what it will spent its resources, it seems quite obvious that the expenditures for the single
categories are correlated with each other. This correlation between disturbances from different
equations at a given time is known as contemporaneous correlation (Judge et al. (1988: 443 ff.)).
The method of seemingly unrelated regression estimation (SURE) controls for this
contemporaneous correlation and provides efficient estimates (going back to Zellner (1962)). It isalso applicable in the given panel data framework. Therefore, we consider the following structural
SURE model with 10 equations to test for the impact of the political variables:
Δlog Public expenditure category ijt = Σk αk Political variable ikt + Σl βl Δlog X ilt + ut
with j = 1,…, 10; k = 1,…,5; l=1,…,8 (1)
9
Note that Spain, Portugal and Greece became democracies in the mid seventies. This explains the somewhatsmaller sample size of our political variables.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 10/47
Niklas Potrafke
10
where the dependent variable Δlog Public expenditure category ijt denotes the first differences of
the change in expenditure category j as a share of GDP. We distinguish between nine expenditure
categories and also consider total spending as a further equation, so that there are ten equations
in total. Political variableikt introduces the political variables, on which this study focuses. The
next paragraph describes its coding in some more detail. Σl Δlog X ilt contains eight exogenous
control variables as well as a constant. We follow the related studies to include: The first
differences of the change in total population, in the share of the young population (aged 14 and
below as a share of total population), in the share of the elderly population (aged 65 and above as
a share of total population), in per capita income (in real terms), in trade as share of GDP, in
prices of public consumption (Tridimas (2001)), as well as in the unemployment rate. Lastly we
also include the lagged dependent variable.
Political variableijt is in the centre of our analysis. We distinguish between a variable controlling
for the effect of election years, the ideological party composition of the governments, the number
of coalition partners, the fact if the respective governments had a majority in parliament or not
(minority government) as well as an interaction term of the ideology variable and the minority
government dummy.
The variable Electionit takes the exact timing of the elections into account. Following Franzese(2000) and the application of De Haan and Mink (2005), it is calculated as
Electionit = [(M-1) + d/D]/12
where M is the month of the election, d is the day of the election and D is the number of days in
that month. In non-election years, its values are set to zero. Therefore, we directly control for
fluctuations and the fact that the election dates differ between as well as in the single countries.
An important challenge for the partisan test in an OECD panel is the heterogeneity of the parties
and parliamentary systems in the single states. Hence the question comes up what kind of
government could be labelled leftwing or rightwing – especially when there are more than two
parties in government with different ideological roots. Bjørnskov (2005a: 4) concludes on this
issue: “Political ideology is a potentially complex feature yet operationalizing it as a
unidimensional construct measurable in a left-to-right scheme hence need not necessarily entail
any sizeable loss of information or sophistication when connected to real political outcomes.”
Researchers often use the index by Budge et al. (1993) and updated by Woldendorp et al. (1998)
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 11/47
Niklas Potrafke
11
and (2000) as a measure of the governments’ ideological positions.10 It locates the cabinet on a
left-right scale with values between 1 and 5. It takes the value 1 if the share of rightwing parties in
terms of seats in government and their supporting parties in parliament is larger than 2/3, 2 if it
is between 1/3 and 2/3. The index is 3 in a balanced situation if the share of centre parties is 50
per cent, or if the leftwing and rightwing parties form a government together not dominated by
one or the other side. Corresponding to the first two cases it takes the values 4 and 5 by a
dominance of the leftist parties likewise defined. Following this procedure, Potrafke (2007a)
applies an ideological index for a group of 20 OECD countries in the period from 1980 to 2003.
It is extended for the current analysis. Consequently, we get a uniform quantitative measure.
Finally, we label years in which the government changed corresponding to the one that was in
office for the longer period, e.g., when a rightwing government followed a leftist in August, we
label this year as leftist. Note that the coding of the ideology variable implies a positive impact on
public expenditures favoured by the left.
Moreover, the type of government is tested by two variables. Roubini and Sachs (1989)
constructed an index of power dispersion which distinguishes between the number of coalition
partners as well as if the government was a minority government. Unfortunately, this procedure
mixes the quantitative feature of the number of parties in the coalition with a qualitative feature,
namely if this government has a majority in parliament or not (Edin and Ohlsson (1991), DeHaan and Sturm (1994, 1997)). Therefore, we first install a variable controlling for the number of
parties in government. It ranges from 0 (no coalition) to 2 (huge coalition):
0 one-party majority parliamentary government;
1 coalition parliamentary government with two-to-three coalition partners;
2 coalition parliamentary government with four or more coalition partners.
Further we use a simple dummy variable to control for the impact of minority governments. It
takes on the value “1” when the government does not have a majority in parliament and zero
otherwise.
Moreover, the impact of the ideological government orientation might be mitigated when the
governments do not have a majority in parliament. They are dependent on the goodwill of other
10 Note that Bjørnskov (2005b) recently introduced a further cross-country indicator of political ideology for the
period from 1976 to 2000. It refers to the data base by Beck et al. (2001) and places the parties on a discrete left-to-right scale where left parties are assigned the value -1, center parties 0 and right parties 1. Lastly the value of thegovernment ideology g t is the weighted sum of the respective ideologies of the three largest parties in government.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 12/47
Niklas Potrafke
12
parties in parliament and therefore cannot implement their pure ideology. We control for this
effect by including the interaction term of our ideology variable and the minority government
dummy.
Lastly, we comment on the specialties regarding the estimation procedure. Taking first
differences of the levels due to stationarity reasons eliminates time-invariant fixed country effects.
Hence, the common least squares dummy variable estimator (fixed-effects) could be useless. But
there could also be linear time trends in each country. Hence, after first differencing, they appear
as time-invariant country effects in the model. Thus, it might also be sensible to apply fixed
country effects on the model in growth rates. Note further, that the constant included in our
regressions does not contradict estimating fixed country effects. The fixed effects are
implemented by simple country dummies and there is one dummy excluded in each regression to
avoid multicollinearity problems. However, in the context of a dynamic model specification using
the lagged dependent variable as regressor, the common fixed-effect estimator might be biased.
We cannot solve this problem in the SURE framework. But there are panel estimators that
correct for this problem. As Behr (2003) states, the estimators taking into account the resulting
bias can be grouped broadly into the class of instrumental estimators and the class of direct bias
corrected estimators. In accordance with large sample properties of the GMM methods, e.g., the
estimator proposed by Arellano and Bond (1991) will be biased in the current framework with N= 23 or less. That is why bias corrected estimators might be a good choice. Bruno (2005)
presents a bias corrected least squares dummy variable estimator for dynamic panel data models
with small N. Hence we will also apply Bruno’s (2005) bias corrected estimator when fixed
country effects are present as a reference. This procedure is less efficient than SURE because it
does not correct for the contemporaneous correlation. We therefore face the tradeoffs between
potential bias and efficiency similar to the related literature.
5. Results
5.1 The period from 1970 to 1989
The distinction between the two subperiods from 1970 to 1989 and from 1990 to 2005 is first
caused by the availability of the data (section 3). However, as regards content, there is also an
important reason distinguishing between the time before 1990 and afterwards. This is highly
motivated by historical events. In 1990, there was the fall of the “Iron Curtain”; the end of the
“Cold War” arose in these years. Garret (1998: 1) states that “…one should be recitent to
conclude differently about the 1990s because of the highly idiosyncratic nature of the decade in
Europe”. In addition, the nineties were also claimed as the end of the welfare states. Hence, the
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 13/47
Niklas Potrafke
13
historical background provides good reasons to examine respective sub-periods. Previous
research has shown empirically that there were differences between the 80ies and the 90ies (e.g.,
Kittel and Obinger (2003), Potrafke (2007a)).
Tables 2 and 3 about here
The estimation results in Table 2 refer to a SURE model with fixed country effects for the period
from 1970 to 1989. We first checked a fixed country effect versus a pooled regression. An F-Test
that all the fixed country effects are zero could be rejected at a 1 percent significance level.
Moreover estimating SURE and OLS would be equal, if there would be no contemporaneous
correlation between the single equations (or completely the same regressors in every equation are
used)11. The Breusch-Pagan-Tests for no contemporaneous correlation can be strongly rejected.
Hence, there are strong efficiency gains from using SURE in the considered model. Of the
contents, the expenditures in one category are dependent on the expenditures in the other
categories, as expected. F-Tests on the control variables point out that they are all jointly
significant at convenient levels – except for the elderly share. Table 3 also shows the regression
results for the single equations using Brunos (2005) dynamic bias corrected estimator.12
Tables 2 and 3 point out that politicians increased spending for “Health” and “Education” as well as for the overall public sector in election years. Hence we would conclude that they behaved
opportunistically and focused on the interests of the median voter to become re-elected. But the
impact of the electoral effects might depend on the fact if the election dates are predetermined;
that is if the particular elections were part of the regular electoral cycle or if they were irregular.
Shi and Svensson (2006) examine political budget cycles and point out that the election timing
might be endogenous. They argue that both timing of elections and fiscal policies could be
influenced by a number of variables, which are not included in the regression. This might cause
omitted variable bias. Mitigating this potential problem Shi and Svensson (2006) provide a
sensitivity analysis focusing on elections whose timing is predetermined relative to current fiscal
policies. Their results do not change for their whole sample but for developed countries. Brender
and Drazen (2005) argue that there are two conceptual problems with Shi and Svensson’s (2006)
presumption. First, it may not be that obvious to distinguish between systems in which electoral
dates are fixed and systems where early elections may be called. In the same manner, early
11 Note that in the current model the equations differ by the lagged dependent variable.12
Bruno’s (2005) estimator can be used in relation to different intitial dynamic panel data estimators. The currentresults refer to the Arellano-Bond (1991) estimator as initial one and standard errors are bootstrapped using 200repetitions.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 14/47
Niklas Potrafke
14
elections seem to be the rule rather than the exception. Furthermore, there are countries where
the government may call early elections, but rarely does. Second, “there is no clear theoretical
presumption about whether fiscal manipulation will be stronger or weaker when election dates
are effectively predetermined” (Brender and Drazen (2005: 1282)). However, Brender and
Drazen (2005) also control for predetermined elections and find that results are not much
affected. In conclusion, as regards contents, distinguishing between regular and irregular elections
seems very reasonable. Thus, we will also consider it in the current paper. We applied Shi and
Svensson’s (2006) data on the predetermined and endogenous election years from 1975 to 1995.13
Furthermore we extend the respective data from 1970 to 197414 and from 1996 to 2005 using the
Political Handbooks, following Shi and Svensson’s (2006: 1374) rules. An election is classified to
be predetermined if either (i) the election is held on the fixed date (year) specified by the
constitution; or (ii) the election occurs in the last year of a constitutionally fixed term for the
legislature; or (iii) the election is announced at least a year in advance. Overall, we detect 63.8
percent of the elections as pre-determined.
Tables 4 and 5 about here
Tables 4 and 5 point out that the conclusion on the electoral cycle was spurious. In fact, the
positive effect of election years on the overall public sector size, expenditures on health andeducation were driven by endogenous elections. The pre-determined election year variables
turned to be insignificant at convenient significance levels. It is implausible that politicians could
exploit the endogeneous elections for short-dated spending policies. They might not pass an
additional budget when early elections are announced. Thus, we cannot conclude that there is
evidence for an opportunistic behaviour of politicians in the period from 1970 to 1989.
Regarding the interpretation of the other policy variables we will refer to the results given in
Tables 4 and 5. The results report that leftist governments did not significantly extend the overall
public sector but allocated public expenditures differently compared to rightwing governments
till the end of the 80ies. They set other priorities. Note that the political effects are somewhat
weaker in the regressions of the single equations. As expected, they increased money for “Public
services” and therefore afford more publicly provided goods than rightwing governments.
Moreover, they disbursed more for “Transport and communications” (SURE), “Housing” and
13
We thank Min Shi and Jakob Svensson for providing their data. 14
Note that even the Political Handbook from 1976 does not exactly identify the general elections in Finland and Japan 1972 as well as the election in Canada in 1974 as endogenous. They are highly expected to be endogenous, buttheoretically, they could have been announced one year in advance. Then we would have to label them aspredetermined. However, this would not affect our inferences.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 15/47
Niklas Potrafke
15
“Health”. Thereby leftist governments seemed to gratify their clientele. However, we expected
leftist governments to spend less for “Defence” and more for “Social welfare”. The results do
not fulfil these prospects. But they might also reflect the fact that the influenced expenditures are
more elastic in the short run and not subject to long run contracts like, e.g., defence.
Moreover, the ideology effects were mitigated when the respective governments did not have a
majority in parliament. Therefore the respective marginal effects of the ideology variables and
their significances have to be interpreted conditional on the interaction with the minority
government dummy (see Friedrich (1982)). In principle, there are two sensible ways to evaluate
the marginal effects. Dreher and Gassebner (2007) suggest evaluating them at the minimum as
well as the maximum of the interacted variable. This procedure also emerges as sensible in our
case as the minority government variable is a dummy facing the values 0 and 1. Hence in case of
a majority government the marginal effect just equals the coefficient of the ideology variable
itself. Its meaning is that a corresponding increase of the ideology variable by one point – say
from 3 (leftist and rightwing parties in government) to 4 (leftwing government) – would increase
the growth rate of public expenditures for “Public services” by 1.5, the one for “Transport and
communications” by 1.7, the one for “Housing” by 3.5 and the one for “Health” by 0.9 percent
(SURE model in Table 4). We further calculated the marginal effects for the case when the
minority variable is 1. Then, all the marginal effects are statistically insignificant. Thus our resultsimply that the hypothesis that parties matter is only fulfilled when they had a majority in
parliament. Without a majority, they were not able to implement their ideological preferences and
to set respective priorities regarding the budget composition. This result confirms an extremely
plausible expectation. In principle, the marginal effects could also be calculated on an average
level of the variable the one of interest is interacted with, but this does not seem sensible
examining the interaction with a simple dummy variable.
Moreover, the results do not allow drawing any conclusions regarding the impact of the size of
coalition variable. It is insignificant in every single equation. Referring to the SURE model, we
further applied F-Tests on the political variables checking their joint significance. F-Tests are
important because they consider the correlation structure between the single parameters and it
could be that all the single coefficients are insignificant in the single equations but jointly
significant.15 Table 6 provides the results. They report that the ideology and minority government
variables are jointly significantly different from zero at 5 or rather 10 percent level.
15
Geometrically, the confidence intervals of the single parameters can be drawn as line segments, whereas the jointconfidence intervals (confidence region) can look like an ellipse. (See, e.g., Judge et al. (1988): 244 ff.).
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 16/47
Niklas Potrafke
16
Table 6 about here
Our results fit the previous literature in the sense that leftist governments did expansionary
policies compared to rightwing governments till the end of the 80ies. As Potrafke (2007b) we
find higher expenditures for “Health”. Furthermore, the current results do not contradict the
recent literature on social expenditures (e.g., Kittel and Obinger (2003) and Potrafke (2007a))
under leftist governments in this period, because our current results do not report a significant
positive impact of leftist governments on “Social welfare”. Note that social expenditures include
spending for health and also housing, for which we got positive effects in our current regressions.
5.2 The period from 1990 to 2005
Table 7 provides the results relating to the panel from 1990 to 2005. They indeed differ from the
results referring to the period from 1970 to 1989. Policy had weaker impacts. In contrast to the
previous model, the F-Tests do not reject the hypothesis that fixed country effects are zero.
Hence we estimate a SURE model with a common constant only. There cannot be a Nickel-bias.
F-Tests on the control variables point out that the three population variables are jointly
insignificant. However, we follow the related literature and keep them in the model. As before,
we first refer to the results with the simple election year variable (Table 7). Politicians seemed toincrease spending for “Health” before elections. There are no further election year effects. Table
8 shows the respective results with predetermined and endogenous elections and points out that
the electoral effect on public health expenditures was only due to endogenous elections. This
finding perfectly corresponds with Potrafke (2007b). Moreover, we detect different spending
preferences of leftist and rightwing governments. At first, parties did not affect the overall
“Public sector size”. Then the results show higher spending for “Cultural affairs” and
“Education” and lower spending for “Social welfare” under leftist governments in the period
from 1990 to 2005. These effects only hold when the respective government had a majority in
parliament. We again calculate marginal effects which are statistically insignificant when the
minority dummy is equal to one. Higher spending for “Education” under leftist governments
fulfils our expectations whereas lower spending for “Social welfare” does not. Also previous
research (e.g. Kittel and Obinger 2003, Potrafke 2007a) has shown that leftist governments did
not implement expansionary policies in the nineties, however, negative impacts are not in line
with the implications of the partisan approach. Moreover, the results do not support the
expectation of higher expenditures for “Housing” and “Health” under leftist and also higher
spending for “Defence” under rightwing governments. The size of coalition did not influence the
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 17/47
Niklas Potrafke
17
budget composition specifically. The results of the F-Tests on the political variables in Table 9
point out that there are no joint significant effects.
The results do not provide evidence for a policy change regarding a particular spending category.
In other words, we do not find statistical significant effects with different size or even sign of a
variable comparing the regressions from 1970 to 1989 and the ones from 1990 to 2005.
Tables 7, 8 and 9 about here
5.3 Robustness of the results
The robustness of the results must be checked in several ways. At first, we will discuss another
econometric specification using somewhat different control variables. The results of Tables 10,
11 and 12 refer to regressions in which we changed three different features. First we also include
the sum of the other expenditures as explanatory variable ( ∑i Expenditure Category i≠j ). The
expenditures for category j must be excluded to avoid endogeneity problems. Hence, the model
controls for the general spending behaviour and implied allocation effects in each equation.
Second, the previous regressions included trade as a share of GDP as it is common in the related
literature. In principle, this could cause endogeneity problems because the dependent variable is
also measured as a share of GDP. Therefore we apply a different measure for trade that alsotakes into account the relative size of the respective country or rather economy by including trade
per capita (sum of imports and exports divided by population). Third, the demographic change
could also be considered by the share of the working population. We therefore replace the shares
of the young and the old by population aged between 15 and 64 as share of the population.16
Tables 10, 11 and 12 about here
Tables 10 and 11 report the robustness of our results and inferences regarding the political
effects on the budget composition in the period from 1970 to 1989. F-Tests of the political
variables also support our findings. The negative impact of leftist governments on “Social
welfare” in the period from 1990 to 2005 turned slightly insignificant (Table 12).
Moreover, we replaced trade as measure of openness or rather globalization by the KOF index of
globalization (Dreher (2006a) and Dreher et al. (2008)). The inclusion of the overall as well as the
16 We choose this variable instead of the dependency ratio which describes the ratio between those of working age
and those of non-working due to data availability for all the countries from 1970 to 2005.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 18/47
Niklas Potrafke
18
three sub indices does not affect our inferences regarding the political variables. Interestingly, we
find negative impacts of the KOF index of economic globalization on the budget composition
from 1970 to 1989.
General government debt might also be an interesting explanatory variable. However, there is the
problem of data availability for the time period from 1970 to 1989 in an OECD panel. The
inclusion of general government debt as a share of GDP results in a regression with 209
observations. In this sample, the positive impact of leftist governments on expenditures for
“Transport and communications” and “Health” turns to be statistically insignificant. Instead,
there is a statistical negative impact of leftist governments on “Education”. Regarding the panel
from 1990 to 2005, the inclusion of public debt does not change our inferences at all. Leftist
governments still affect spending for “Cultural affairs”, “Education” and “Social welfare” as
before.
The regressions in Tables 2, 3, 4, 5, 10 and 11 referring to the period from 1970 to 1989 only
cover 22 OECD countries. Germany is excluded because there are no data on public
consumption prices available till the beginnings of the nineties. Hence we were running
regressions using consumer prices instead of public consumption prices because there are also
data for Germany available. These regressions show that our results presented above are notsensitive to the exclusion of Germany. They further point out that education was a polarizing
political issue in Germany in the seventies and eighties in the sense that leftist governments spent
more for education than rightwing governments. These empirical results indeed fit anecdotal
evidence on the spending preferences of the Social Democrats ruling in Germany from 1969 to
1982.
The differences between the two data sets due to their number of countries covered as well as the
measurement of the single expenditure categories (recoding) might be a caveat against our
conclusion on a policy change with the beginnings of the nineties. Therefore we estimate two
further models countervailing this appeal. First, we only focus on the same 13 countries for the
period from 1970 to 1989 for which we also have data for the period from 1990 to 2005. In this
sample, the policy effects are indeed weaker. Note for example, that especially health is a
polarizing political issue only in some countries like e.g. Australia and New Zealand (Potrafke
(2007b)). But these two are not included in the 13 country panel. Hence the inferences from the
panel referring to only 13 countries might be somewhat selective. Second, we were also running
regressions using the data from Sanz and Velázquez (2007) and Gemmell et al. (2007) for the
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 19/47
Niklas Potrafke
19
period from 1970 to 1997. In this scenario, the policy impacts are indeed weaker than focusing
on the 70ies and 80ies. These findings strongly support a policy change with the beginnings of
the nineties. However, the question remains if there was a structural shift that also affected the
parties and therefore our political variables itself. In other words, we simply applied a static
ideology variable which does not consider that also parties change in time elapsed. They simply
might have changed their preferences. Potrafke (2007c) develops a procedure for the
construction of a dynamic ideology index and applies it to German data. It exactly controls for
this shortcoming.
As it is common in the literature, we check for the sensitivity of the results to individual
countries. Therefore, we rerun the regressions excluding one country at a time. Some of the
results are sensitive to the inclusion of particular countries. Regarding the first data set, the
impact of the ideology variables strongly declines when New Zealand is excluded. Also, excluding
Australia turns down the impact of the ideology variable on expenditures on “Health”. Leftist
governments in the USA spent much more for “Public services” than rightwing governments.
Denmark, Spain and Sweden are countries in which minority governments were in power. Not
surprisingly, the impacts of minority governments are strongly driven by these countries.
Regarding the OECD panel from 1990 to 2005, excluding Belgium, Japan, the UK and the USA
weakens the impacts of the ideology variables. Interestingly, the USA mainly drives the negativeeffect of leftist government on “Social protection”.
6. Conclusion
This paper showed how political effects determined the allocation of public expenditures (general
government) in OECD countries from 1970 to 2005. Examining policy impacts on the budget
composition requires a detailed expenditure composition because the broader the classification
the higher the probability of compensating policy effects. Therefore, we focused on the COFOG
classification with 9 different subcategories. As there is no comprehensive data set available yet,
we analyzed two different ones: The data set collected by Sanz and Velázquez (2007) and
Gemmell et al. (2007) for the period from 1970 to 1989 as well as current data from the OECD
for the period from 1990 to 2005. Our results demonstrated that there were strong policy effects
in the first period. In line with the expectations, leftist governments disbursed more than
rightwing governments for “Public services”, “Transport and communications”, “Housing” and
“Health”. But interestingly, we did not detect a statistically significant effect on total government
spending. This finding highly fortified the aim of the current paper examining the allocation of
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 20/47
Niklas Potrafke
20
public expenditures because there might even be compensating effects between the categories.
Moreover, we controlled for the interaction between government ideology and minority
governments. Our results clearly demonstrated that parties did only matter when they had a
majority in parliament. Lastly, applying the distinction between predetermined and endogenous
election year effects introduced by Shi and Svensson (2006), we indeed detected the electoral
effects as spurious.
Effects were different using the current OECD data set from 1990 to 2005: The political impacts
of the budget composition became weaker. Furthermore, parties set other priorities in the sense
that leftist governments increased spending for “Cultural affairs” and “Education” and even
decreased for “Social welfare”. In conclusion, our analysis pointed out that there was a policy
shift. This change was not only due to the different data sets. However, more detailed inferences
would require a comprehensive data set.
Therefore the results of the current paper provide two important implications for future research.
First, dynamic policy variables especially ideology indices should be developed and applied in
empirical research considering the changes in time elapsed. Our results are perfectly in line with
the previous research in the sense that we detected a policy change with the beginning of the
nineties. Leftist governments did not do expansionary policies any more. However, it would beinteresting to examine if these policy changes are still present when the political variables also
take into account that parties change in time elapsed. Second, a comprehensive data set on the
budget composition (COFOG) seems to be very helpful for future research. Further research
might composite and examine a unique data set on the budget composition. Then we could
determine if our analysis of political effects as well as the results in the very recent literature on
the allocation of public expenditures (Dreher et al. (2007), Sanz and Velázquez (2007), Gemmell
et al. (2007) and Shelton (2007)) hold in a unique and extended data set. Moreover, one could
determine the most important driving factors on the allocation of public expenditures and further
test for their interactions.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 21/47
Niklas Potrafke
21
Appendix: Panel unit root tests
The following tables report the results of the panel unit root tests on the government
expenditures as a share of GDP as well as the tests on the respective series of population, the
elderly and young share, per capita income, trade as a share of GDP, prices of public
consumption, the unemployment rate, the (population) share aged 15-64 and trade per capita for
the panel from 1970 to 1989 as well as for the panel from 1990 to 2005. They refer to the test on
the first differences. We applied the Levin et al. (2002), Im et al. (2003) and the Fisher tests
referring to Maddala and Wu (1999) and Choi (2001). In contrast, the Hadri (2000) test uses the
null hypothesis that there is no unit root. Breitung and Pesaran (2005) provide a detailed
description of the recent panel unit root tests. The results were obtained using Eviews 5.1. In
comparison to STATA 9.1, Eviews 5.1. allows to apply the respective tests on unbalanced panels,
it considers an automatic lag length selection by the use of Information Criteria and also contains
the Breitung (2000) test. Regarding the first three tests listed in the table, maximum lag lengths
are automatically selected based on the Schwarz Information Criterion. The remaining two tests
use the Bartlett kernel for the Newey-West bandwidth selection. The probabilities for the Fisher
tests are computed using an asymptotic Chi-square distribution. All other tests assume asymptotic
normality (See Hartwig (2007: Appendix)).
Tables 13 to 20 about here
Tables 13 to 20 report the results of different unit root tests and demonstrate that we can always
reject the null hypotheses of a unit root in first differences except for the elderly and young share
and prices of public consumption17 However, the Hadri tests indicate that regarding most of the
series, there are still unit roots in first differences. Overall, we conclude from these tests that the
time series in first differences are stationary. However, at first, we applied the respective tests on
the level series, of course. At this stage, the inclusion of a deterministic trend in the test
regression emerges as crucial. There are pros and cons to include linear time trends testing for
unit roots in time series in levels. Statistically, the plots of the series demonstrate their rise in time
elapsed. Furthermore including a trend is necessary so that the process can mirror the data under
the alternative hypothesis. This definitely demands the inclusion of linear time trends. As regards
content, one might argue that linear time trends must be excluded because the respective
variables cannot steadily increase, e.g., the rise of public spending could not be infinite. As a
17Note that when the share of a respective category is zero in a single country (e.g., “Defence” or “Cultural Affairs”),then Eviews 5.1. does not take into account this country because the logarithm of zero is not defined. In contrast, weindeed consider these observations in our regressions by setting the first difference of the logarithm also equal tozero.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 22/47
Niklas Potrafke
22
result, we include a linear trend in the test regression in levels. Moreover, taking first differences
eliminates linear time trends so that we did not include them in the test regressions in first
differences. Overall, the tests on the levels mostly indicate that the respective series are non-
stationary in levels. Therefore, we estimate the models in first differences. We take first
differences of all variables for consistency reasons, running the risk of over- or under-
differentiation of single variables.18
Finally, the applied battery of panel unit root tests is standard and fits the ones in the existing
literature. More sophisticated tests (as described in Breitung and Pesaran (2005)) do not seem
necessary in the current paper. In particular, so called second generation panel unit root tests
might be applied, that take into account the fact that the time series might not be independent
across i, but contemporaneously correlated. However, Breitung and Pesaran (2005: 19) claim that
“the literature on modelling of cross section dependence in large panels is still developing”. The
named tests suggest that estimating in first differences will be the best specification. Moreover,
one might put doubt on the inference of any unit root test due to the relatively small number of
observations in the sample. In this case, researchers might refer to their empirical intuition
specifying the model. We indeed strongly believe that using first differences is the most
appropriate way in the current framework.
18Finally, the applied battery of panel unit root tests is standard and fits the ones in the existing literature. Moresophisticated tests (as described in Breitung and Pesaran (2005)) do not seem necessary in the current paper. Inparticular, so called second generation panel unit root tests might be applied, that take into account the fact that thetime series might not be independent across i, but contemporaneously correlated. However, Breitung and Pesaran(2005: 19) claim that “the literature on modelling of cross section dependence in large panels is still developing”. Thenamed tests suggest that estimating in first differences will be the best specification. Moreover, one might put doubt
on the inference of any unit root test due to the relatively small number of observations in the sample. In this case,researchers might refer to their empirical intuition specifying the model. We indeed strongly believe that using firstdifferences is the most appropriate way in the current framework.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 23/47
Niklas Potrafke
23
References
Alesina, A. 1987. Macroeconomic policy in a two-party system as a repeated game. The Quarterly Journal of Economics 102 (3), 651-678.
Alesina, A., Nouriel, R., Cohen, G. D. 1997. Political cycles and the macroeconomy. Cambridge: The MIT Press.
Alt, J. E., Lassen, D. D. 2006. Fiscal transparency, political parties, and debt in OECD countries.European Economic Review 50, 1403-1439.
Arellano, M., Bond, S. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidenceand an Application to Employment Equations. Review of Economic Studies 58, 277–297.
Bawn, K. 1999. Money and majorities in the Federal Republic of Germany: Evidence from a vetoplayers model of government spending. American Journal of Political Science 43, 707-736.
Beck, T.,Clarke, G., Groff, A., Keefer, P., Walsh, P. 2001. New tools in comparative politicaleconomy: The database of political institutions. World Bank Economic Review 15, 165-176.
Behr, A. 2003. A comparison of dynamic panel data estimators: Monte Carlo evidence and anapplication to the investment function. Discussion paper 05/03, Economic Research Centre ofthe Deutsche Bundesbank.
Belke, A. 1996. Politische Konjunkturzyklen in Theorie und Empirie. Tübingen: Mohr-Siebeck Verlag.
Belke, A. 1997a. Politischer Konjunkturzyklus – Kritische Anmerkungen zum Erklärungsbeitragdes Parteigänger-Ansatzes. Ifo-Studien 3/1997 43, 311-346.
Belke, A. 1997b. Regierungsideologie, Geldpolitik und Persistenz der Arbeitslosigkeit, Volkswirtschaftliche Diskussionsbeiträge der Ruhr-Universität Bochum, Nr. 7, Bochum.
Belke, A. 1997c. Tests politischer Konjunkturzyklen vom Partisan-Typ: Problematik undempirische Evidenz im Vergleich Deutschlands und der Vereinigten Staaten, RWI-Mitteilungen47, 181-214.
Belke, A. 2000. Partisan political business cycles in the German labour market? Empirical tests inthe light of the Lucas-critique, Public Choice 104, 225-283.
Bjørnskov, C. 2005a. Political ideology and economic freedom. Working Paper 05-8, Departmentof Economics, Aarhus School of Business.
Bjørnskov, C. 2005b. Does political ideology affect economic growth. Public Choice 123, 133-146.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 24/47
Niklas Potrafke
24
Boix, C. 1997. Political parties and the supply side of the economy: The provision of physical andhuman capital in advanced economies, 1960-90. American Journal of Political Science 41 (3),814-845.
Bräuninger, T. 2005. A partisan model of government expenditure. Public Choice. 125, 409-429.
Breitung, J. 2000. The local power of some unit root tests for panel data. In: Baltagi, B. (Eds.), Advances in Econometrics, Vol. 15: Nonstationary panels, panel cointegration, and dynamicpanels. JAI Press: Amsterdam; 1990, 161-178.
Breitung, J., Pesaran, M. H. 2005. Unit Roots and cointegration in panels. CESifo WorkingPaper No. 1565.
Brender, A., Drazen, A. 2005. Political budget cycles in new versus established democracies. Journal of Monetary Economics 52, 1271-1295.
Bruno, G. S. F. 2005. Estimation and inference in dynamic unbalanced panel data models with asmall number of individuals. CESPRI WP n. 165, Università Bocconi-CESPRI, Milan.
Budge, I., Keman, H., Woldendorp, J. 1993. Political data 1945-1990. Party government in 20democracies. European Journal of Political Research 24 (1), 1-119.
Choi, I. 2001. Unit root tests for panel data. Journal of International Money and Finance 20 (2),249-272.
Correa, H., Kim, J.-W. 1992. A causal analysis of defence expenditures of the USA and theUSSR. Journal of Peace Research 29 (2), 161-174.
De Haan, J., Mink, M. 2005. Has the Stability and Growth Pact impeded political budget cycles inthe European Union? CESifo Working Paper No. 1532.
De Haan, J., Sturm, J.-E. 1994. Political and institutional determinants of fiscal policy in theEuropean Community. Public Choice 80, 157-172.
De Haan, J., Sturm, J.-E. 1997. Political and economic determinants of OECD budget deficitsand government expenditures: A reinvestigation. European Journal of Political Economy 13, 739-750.
Dixit, A., Londregan, J. 1996. The determinants of success of special interests in redistributivepolitics. Journal of Politics 58, 1132-1155.
Drazen, A., Eslava, E. 2005. Electoral manipulation via voter-friendly spending: Theory andevidence. Working Paper. Tel Aviv University.
Dreher, A. 2006a. Does globalization affect growth? Evidence from a new index of globalization. Applied Economics 38 (10), 1091-1110.
Dreher, A. 2006b. The influence of gloablization on taxes and social policy – an empiricalanalysis of OECD countries. European Journal of Political Economy 22 (1), 179-201
Dreher, A., Gassebner, M. 2007. Greasing the wheels of entrepreneurship? The impact ofregulations and corruption on firm entry. CESifo Working Paper No 2013.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 25/47
Niklas Potrafke
25
Dreher, A., Gaston, N., Martens, P. 2008. Measuring globalization – Understanding its causesand consequences. Springer: Berlin, forthcoming.
Dreher, A., Sturm, J.-E., Ursprung, W. 2007. The impact of globalization on the composition of
government expenditures: Evidence from panel data. Public Choice. forthcoming.
Edin, P.-A., Ohlsson, H. 1991. Political determinants of budget deficits: Coalition effects versusminority effects. European Economic Review 35, 1597-1603.
Franzese, R. 2000. Electoral and partisan manipulation of public debt in developed democracies,1956-1990”, in: Strauch, R., von Hagen, J. (Eds.), Institutions, politics and fiscal policy.Dordrecht: Kluwer Academic Press, pp. 61-83.
Friedrich, R. J. 1982. In defence of multiplicative terms in multiple regression equations. American Journal of Political Science 26, 797-833.
Galli, E., Rossi, S. 2002. Political budget cycles: The case of the Western German Länder. PublicChoice 110, 283-303.
Garrett, G. 1998. Partisan politics in the global economy. Cambridge: Cambridge UniversityPress.
Gemmell, N., Kneller, R., Sanz, I. 2007. Foreign investment, international trade and the size andstructure of public expenditures. European Journal of Political Economy. DOI:10.1016/j.ejpoleco.2007.06.004
Hadri, K 2000. Testing for stationarity in heterogeneous panel data. Econometrics Journal 3 (2),148-161.
Hartwig, J. K. 2007. What drives health care expenditure? – Baumol’s model of ‘unbalancedgrowth’ revisited. Journal of Health Economics, forthcoming.
Hibbs, D. A. Jr. 1977. Political parties and macroeconomic policy. The American PoliticalScience Review 71, 1467-1487.
Hicks, A. M., Swank, D. 1992. Politics, institutions and welfare spending in industrialized
democracies, 1960-82. The American Political Science Review 86 (3), 658-674.
Im, K. S., Pesaran, M. H., Shin, Y. 2003. Testing for roots in heterogenous panels. Journal ofEconometrics 115 (1), 53-74.
Immergut, E. M. 1992. Health politics – Interests and institutions in Western Europe. CambridgeUniversity Press: Cambridge.
Judge, G. G., Hill, R. C., Grifftths, W. E., Lütkepohl, H., Lee, T.-S. 1988. Introduction to thetheory and practice of econometrics. New York a.o.: John Wiley & Sons.
Kittel, B., Obinger, H. 2003. Politicial parties, institutions, and the dynamics of social expenditurein times of austerity. Journal of European Public Policy 10 (1), 20-45.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 26/47
Niklas Potrafke
26
König, T., Tröger, V. 2005. Budgetary politics and veto players. Swiss Political Science Review 11(4), 47-75.
Levin, A., Lin, C. F., Chu, C. 2002. Unit root tests in panel data: Asymptotic and finite-sample
properties. Journal of Econometrics 108 (1), 1-24.
Lindbeck, A., Weibull, J. W. 1987. Balanced-budget redistribution as the outcome of politicalcompetition. Public Choice 52, 273-297.
Maddala, G. S., Wu, S. 1999. A comparative study of unit root tests with panel data and a newsimple test. Oxford Bulletin of Economics and Statistics 61, 631-652.
Mueller, D. C. 2003. Public Choice. Cambridge: Cambridge University Press.
Nincic, M., Cusack, T. R. 1979. The political economy of US military spending. Journal of PeaceResearch 16 (2), 101-115.
Nordhaus, W. D. 1975. The political business cycle. The Review of Economic Studies 42 (2),169-190.
Persson, T., Tabellini, G. 2000. Political economics – Explaining economic policy. Cambridge: The MIT Press.
Potrafke, N. 2006. Parties matter in allocating expenditures: Evidence from Germany. DIW
Discussion Papers No. 652.
Potrafke, N. 2007a. Social expenditures as a political cue ball? Empirical evidence from OECDcountries. DIW Discussion Papers No. 676.
Potrafke, N. 2007b. Public health expenditures in OECD countries: Does policy matter?Humboldt University Berlin, mimeo.
Potrafke, N. 2007c. Parties change! The development of a dynamic ideology index and itsconsideration in empirical research. Humboldt University Berlin; mimeo.
Rogoff, K., Sibert, A. 1988. Elections and macroeconomic policy cycles. The Review ofEconomic Studies 55 (1), 1-16.
Roubini, N., Sachs, J. D. 1989. Political and economic determinants of budget deficits in theindustrial democracies. European Economic Review 33, 903-938.
Sanz, I., Velázquez, F. 2007. The role of aging in the growth of government and social welfarespending in the OECD. European Journal of Political Economy. DOI:10.1016/j.ejpoleco.2007.01.003.
Shelton, C. A. 2007. The size and composition of government expenditure. Journal of Public
Economics; DOI:10.1016/jpubeco.2007.01.003.Shi, M., Svensson, J. 2006. Political budget cycles: Do they differ across countries and why?
Journal of Public Economics 90, 1367-1389.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 27/47
Niklas Potrafke
27
Tridimas, G. 2001. The economics and politics of the structure of public expenditure. PublicChoice 106, 299-316.
Tsebelis, G. 1995. Decision making in political systems: Veto players in presidentialism,parliamentarism, multicameralism and multipartyism. British Journal of Political Science 25, 289-
325.
Tsebelis, G. 2002. Veto players – How political institutions work. Princeton: Princeton UniversityPress.
Tsebelis, G., Chang, E. C. C. 2004. Veto players and the structure of budgets in advancedindustrialized countries. European Journal of Political Research 43 (3), 449-476.
Weingast, B. R., Shepsle, K. A., Johnson, C. 1981. The political economy of benefits and costs: Aneoclassical approach to distribution politics. Journal of Political Economy 89, 642-664.
Woldendorp, J., Keman, H., Budge, I. 1998. Party government in 20 democracies: an update
1990-1995. European Journal of Political Research 33, 125-164.
Woldendorp, J., Keman, H., Budge, I. 2000. Party government in 48 democracies 1945-1998:composition, duration, personnel. Dordrecht: Kluwer Academic Publishers.
Zellner, A. 1962. An efficient method of estimating seemingly unrelated regressions and tests ofaggregation bias. Journal of the American Statistical Association 57, 348-368.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 28/47
Niklas Potrafke
28
Table 1. Expected effects of the political determinants on the expenditure categories
Expenditure Category Election year Leftist government Type of government
Public services +Defence –Economic affairs
Transport and communicationsEnvironmental protection +Housing + +Health + +Cultural AffairsEducation + +Social welfare + +
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 29/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 30/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 31/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 32/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 33/47
Niklas Potrafke
Table 6: F-Tests of the political variables. 1970-1989.
Variable F-Statistic P-Value
Election (predetermined) 0.92 0.5153Election (endogenous) 0.90 0.5323Ideology** 1.96 0.0338Size of coalition 0.61 0.8063
Minority Government* 1.67 0.0813Ideology, Ideology*Minority government* 1.46 0.0848Size of coalition, Minority government** 1.15 0.2890
*/**/***: jointly significant at the 0.10/0.05/0.01 level.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 34/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 35/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 36/47
Niklas Potrafke
Table 9: F-Tests of the political variables. 1990-2005.
Variable F-Statistic P-Value
Election (predetermined) 0.30 0.9819Election (endogenous) 0.75 0.6812Ideology 1.45 0.1544Size of coalition 1.06 0.3929Minority Government 0.37 0.9580Ideology, Ideology*Minority government 0.75 0.7702Size of coalition, Minority government 0.91 0.5762
*/**/***: jointly significant at the 0.10/0.05/0.01 level.
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 37/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 38/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 39/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 40/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 41/47
Niklas Potrafke
Table 16: Panel unit root tests. Control variables, second part.. 1970-1989.log (Prices public consump.) log (Unemployment) log (Share aged 15-64) log (Trade per capita)
Ho: Unit root in first diff. Stat. Prob. Obs. Stat. Prob. Obs. Stat. Prob. Obs. Stat. Prob.
Levin, Lin & Chu t* -2.414 0.008 390 -11.145 0.000 378 -1.753 0.040 350 -14.035 0.000
Im, Pesaran & Shin W-stat -1.1153 0.124 390 -8.786 0.000 378 -1.342 0.090 350 -11.345 0.000
ADF-Fisher Chi-square 49.185 0.273 390 169.455 0.000 378 81.518 0.001 350 206.470 0.000
PP-Fisher Chi-square 46.126 0.384 396 168.645 0.000 389 21.756 0.999 414 251.536 0.000
Ho: No unit root in first diff.
Hadri Z-stat 6.379 0.000 418 3.525 0.002 412 6.267 0.030 437 0.879 0.190
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 42/47
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 43/47
Niklas Potrafke
Table 20: Panel unit root tests. Control variables, second part.. 1990-2005.log (Prices public consump.) log (Unemployment) log (Share aged 15-64) log (Trade per capita)
Ho: Unit root in first diff. Stat. Prob. Obs. Stat. Prob. Obs. Stat. Prob. Obs. Stat. Prob.
Levin, Lin & Chu t* -5.401 0.000 205 -5.656 0.000 201 -1.892 0.029 208 -7.567 0.000
Im, Pesaran & Shin W-stat -3.913 0.000 205 -5.190 0.000 201 0.351 0.637 208 -5.840 0.000
ADF-Fisher Chi-square 57.786 0.003 205 75.766 0.000 201 34.069 0.133 208 78.190 0.000
PP-Fisher Chi-square 58.085 0.003 205 55.556 0.006 201 75.706 0.000 208 81.091 0.000
Ho: No unit root in first diff.
Hadri Z-stat 2.938 0.002 206 1.988 0.023 201 7.779 0.030 208 0.564 0.286
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 44/47
Data description and sources
Descriptive Statistics. 1970-1989.
Variable Observations Mean Std. Dev. Min Max Source
Public sector size 460 37.48 8.83 18.43 60.69Sanz and Velázquez (2007)Gemmell et al. (2007)
Public services 460 3.69 1.15 1.06 6.51Sanz and Velázquez (2007)Gemmell et al. (2007)
Defence 460 2.40 1.38 0 8.18Sanz and Velázquez (2007)Gemmell et al. (2007)
Economic Affairs 460 3.41 1.94 1.09 11.30Sanz and Velázquez (2007)Gemmell et al. (2007)
Environmentalal protection 460 2.91 1.24 0.75 7.41Sanz and Velázquez (2007)Gemmell et al. (2007)
Housing 460 1.45 0.95 0.12 4.82Sanz and Velázquez (2007)Gemmell et al. (2007)
Health 460 4.91 1.38 1.47 9.87Sanz and Velázquez (2007)Gemmell et al. (2007)
Cultural affairs 460 0.82 0.48 0.00 2.34Sanz and Velázquez (2007)Gemmell et al. (2007)
Education 460 5.16 1.45 0.90 8.60
Sanz and Velázquez (2007)
Gemmell et al. (2007)
Social welfare 460 12.73 5.96 2.61 28.56Sanz and Velázquez (2007)Gemmell et al. (2007)
Election 446 0.18 0.31 0 1 own collection
Election (predetermined) 446 0.11 0.25 0 0.95 own collection
Election (endogenous) 446 0.08 0.22 0 1 own collection
Ideology 441 2.84 0.89 1 4 own collection
Size of coalition 441 0.74 0.77 0 2 own collection
Minority government 441 0.16 0.37 0 1 own collection
Total population 460 31763.06 49940.52 205.00 246819.00 OECD Health Data (2007)
Young share 460 22.96 3.67 15.83 32.34 Worldbank (2007)
Elderly share 460 12.26 2.25 7.07 17.98 Worldbank (2007)
Share aged 15-64 460 64.78 2.54 57.65 70.38 Worldbank (2007)
Per capita income 459 16216.95 5825.97 4281.56 32133.00 Worldbank (2007)
Trade (as a share of GDP) 458 62.51 35.44 11.25 208.71 Worldbank (2007)
Trade (per capita) 459 81744.79 228720.20 1069.03 1453264.00 Worldbank (2007)
Prices public consumption 440 37.76 23.19 0.17 89.13 OECD
Unemployment rate 441 5.16 4.03 0.00 21.00 OECD Health Data (2007)
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 45/47
Niklas Potrafke
45
Descriptive Statistics. 1990-2005.
Variable Observations Mean Std. Dev. Min Max Source
Public sector size 185 47.57 8.05 31.58 67.08 OECD (2007)
Public services 185 9.23 2.70 4.54 17.61 OECD (2007)
Defence 185 1.74 1.06 0.26 5.64 OECD (2007)
Economic Affairs 185 4.51 1.17 2.21 11.10 OECD (2007)
Environmentalal protection 185 0.62 0.40 0.00 1.82 OECD (2007)
Housing 185 0.92 0.44 0.14 2.82 OECD (2007)
Health 185 6.27 0.73 4.10 7.74 OECD (2007)
Cultural affairs 185 0.99 0.47 0.15 2.20 OECD (2007)
Education 185 5.63 1.12 3.92 8.24 OECD (2007)
Social welfare 185 17.64 5.37 6.60 28.36 OECD (2007)
Election 208 0.15 0.28 0 0.96 own collection
Election (predetermined) 208 0.10 0.23 0 0.92 own collection
Election (endogenous) 208 0.05 0.19 0 0.96 own collection
Ideology 208 2.87 0.84 2 4 own collection
Size of coalition 208 0.99 0.72 0 2 own collectionMinority government 208 0.22 0.41 0 1 own collection
Total population 204 54170.96 74487.09 384.00 296410.00 OECD Health Data (2007)
Young share 208 18.21 2.44 14.00 27.33 Worldbank (2007)
Elderly share 208 15.17 2.03 10.91 19.97 Worldbank (2007)
Share aged 15-64 208 66.62 1.59 61.32 70.25 Worldbank (2007)
Per capita income 208 25619.02 7333.80 13784.08 52182.86 Worldbank (2007)
Trade (as a share of GDP) 205 82.32 58.49 16.11 293.87 Worldbank (2007)
Trade (per capita) 201 102146.00 189078.50 4647.01 969978.20 Worldbank (2007)
Prices public consumption 207 91.86 14.38 47.86 122.77 OECD:Economic Outlook
Unemployment rate 201 7.26 3.51 1.10 16.40 OECD Health Data (2007)
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 46/47
Niklas Potrafke
46
Variable description Variable Description SourcePublic sector size(total spending)
Aggregate government spending excluding interestsas a share of GDP
OECD (2007)
Public Services(sum of General public
services and Publicorder and safety)
1. Executive and legislative organs, financial and fiscalaffairs, external affairs
2.
Foreign economic aid3. General services4. Basic research5. R&D General public services6. General public services n.e.c.7. Public debt transactions8. Transfers of a general character between different levels of
government9. Police services10. Fire-protection services11. Law courts12. Prisons13. R&D Public order and safety
14.
Public order and safety n.e.c.as a share of GDP
OECD (2007)
Defence 1. Military defence2. Civil defence3. Foreign military aid4. R&D Defence5. Defence n.e.c.
as a share of GDP
OECD (2007)
Economic affairs 1. General economic, commercial and labour affairs2. Agriculture, forestry, fishing and hunting3. Fuel and energy4. Mining, manufacturing and construction5. Transport
6.
Communication7. Other industries8. R&D Economic affairs9. Economic affairs n.e.c.
as a share of GDP
OECD (2007)
Environmentalprotection
1. Waste management2. Waste water management3. Pollution abatement4. Protection of biodiversity and landscape5. R&D Environmentalal protection6. Environmentalal protection n.e.c.
as a share of GDP
OECD (2007)
Housing 1. Housing development
2.
Community development3. Water supply4. Street lighting5. R&D Housing and community affairs6. Environmentalal protection n.e.c.
as a share of GDP
OECD (2007)
Health 1. Medical products, appliances and equipment2. Outpatient services3. Hospital services4. Public health services5. R&D health6. Health n.e.c.
as a share of GDP
OECD (2007)
Cultural affairs 1.
Recreational and sporting services2.
Cultural services3. Broadcasting and publishing services4. Religious and other community services
OECD (2007)
7/23/2019 Political Effects on the Allocation of Public Expenditures
http://slidepdf.com/reader/full/political-effects-on-the-allocation-of-public-expenditures 47/47
Niklas Potrafke
5. R&D Recreation, culture and religion6. Recreation, culture and religion n.e.c.
as a share of GDPEducation 1. Pre-primary and primary education
2. Secondary education3. Post-secondary non-tertiary education
4.
Tertiary education5. Education not definable by level6. Subsidiary services to education7. R&D Education8. Education n.e.c.
as a share of GDP
OECD (2007)
Social welfare 1. Sickness and disability2. Old age3. Survivors4. Familiy and children5. Unemployment6. Housing7. Social exclusion n.e.c.
8.
R&D Social protection9.
Social protection n.e.c.as a share of GDP
OECD (2007)
Population Resident population in thousands Worldbank (2007) Young share Population aged 14 and below as a share of total population Worldbank (2007)Elderly share Population aged 65 and above as a share of total population Worldbank (2007)Gross domesticproduct (per capita)
GDP per capita is gross domestic product divided by midyearpopulation. GDP at purchaser's prices is the sum of gross valueadded by all resident producers in the economy plus any producttaxes and minus any subsidies not included in the value of theproducts. It is calculated without making deductions fordepreciation of fabricated assets or for depletion and degradation ofnatural resources. Data are in constant US dollars.
Worldbank (2007)
Trade Imports of goods and services represent the value of all goods andother market services received from the rest of the world. Theyinclude the value of merchandise, freight, insurance, transport,travel, royalties, license fees, and other services, such ascommunication, construction, financial, information, business,personal, and government services. They exclude labor andproperty income (formerly called factor services) as well as transferpayments. Data are in constant local currency.
Exports of goods and services represent the value of all goods andother market services provided to the rest of the world. Theyinclude the value of merchandise, freight, insurance, transport,travel, royalties, license fees, and other services, such ascommunication, construction, financial, information, business,personal, and government services. They exclude labor andproperty income (formerly called factor services) as well as transferpayments. Data are in constant local currency.
Worldbank (2007)
Prices publicconsumption
Deflator, Public consumption, 2000=100 OECD: EconomicOutlook (2007)
Unemployment Rate Total unemployment, % of labor force OECD HealthData Base (2007)
Share aged 15-64 Population aged between 15 and 64 as a share of total population Worldbank (2007)