Original Article
Rail subsidisation in the European Union: An
issue beyond left and right?
Katharina Crössmanna
and Karsten Mauseb,*
a
University of Bremen, Center for Social Policy Research (ZeS), Mary-Somerville-Str. 5, D-28359
Bremen, Germany.
E-mail: kcroessmann@uni-bremen.de.
b
University of Münster, Institute of Political Science (IfPol), Scharnhorststr. 100, D-48151 Münster,
Germany.
E-mail: karsten.mause@uni-muenster.de.
*Corresponding author.
Abstract Despite the considerable amount of public subsidies granted by the member
states of the European Union (EU) to their national railway industries, the determinants of
State aid to this particular sector are still largely unexplored from a cross-country perspective.
Using ofﬁcial subsidy data for a sample of 25 EU countries over the period
1998–2008, this article poses the classic do-parties-matter question in comparative political
economy to examine whether the political allocation of railway subsidies is
completely determined by sector-speciﬁc conditions or whether it leaves room for governments’
partisan preferences. Controlling for other potential politico-economic determinants
(sector size, public/private ownership, intermodal competition and so on), a
multiple regression analysis indicates that government ideology in fact helps explain the
considerable differences between countries with respect to rail subsidisation. Contrary to
the expectations of traditional partisan theory, countries governed by left-wing (bourgeois)
parties showed signiﬁcantly lower (higher) subsidy levels. The article goes deeper
into this regression result and discusses a number of explanations as to how this unexpected
empirical ﬁnding can be explained.
Comparative European Politics advance online publication, 17 February 2014;
doi:10.1057/cep.2014.2
Keywords: subsidies; railways; European Union; political parties; comparative political
economy
Introduction
With the exception of Cyprus and Malta, which ‘have no railway operators’
(European Commission (EC), 2012a, p. 6), there is no member state of the European
© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
www.palgrave-journals.com/cep/
Union (EU) that does not ﬁnancially subsidise its national railway industry. On the
basis of an investigation of ﬁrm-level data for six major train operating companies in
Austria, Denmark, France, Germany, Great Britain and Spain in the ﬁnancial year
2006, a report by the US National Railroad Passenger Corporation (AMTRAK) in
this context notes that ‘[a]lthough some European train operating companies may
report a “proﬁt”, this proﬁt is generated through a large amount of public funding
provided by the European countries’ (AMTRAK, 2008, p. 5). Even with the best of
intentions, and although policymakers may have good reasons for subsidising this
industry (for example, public service considerations, environmental protection,
railway noise mitigation), economists might argue that many of these public cash
injections are actually harmful because of having incentive effects (soft budget
constraints) and possibly causing a kind of subsidy ‘mentality’ or ‘addiction’ (Otáhal
and Pospíšil, 2009).
Although the economic perspective of examining the effects of subsidies on the
efﬁciency of railway undertakings is indeed an interesting one (see, for example,
Oum and Yu, 1994; Friederiszick et al, 2003), this study takes a political science
perspective and explores whether the governmental subsidisation of the national
railway industries in 25 EU member countries over the period 1998–2008 was
inﬂuenced by partisan motives. Although this potential explanatory factor belongs
to the ‘usual suspects’ analysed in comparative studies of industrial subsidies as
a tool of economic policy (see, for example, Obinger and Zohlnhöfer, 2007;
Zahariadis, 2008, 2010; Keman, 2010), so far there has not been a cross-country
study that investigates the issue of whether government ideology and other politicoeconomic
factors matter for the allocation of railway subsidies.
This void is surprising given that these subsidies are a major component of the
total subsidies granted by EU member states to their national economies. For
example, in 2008 railway subsidies amounted to €51.9 billion in the EU-27, whereas
total State aid (minus railways) amounted to €73.9 billion (see EC, 2011).
Considering the 2008 sum of the total State aid plus railway subsidies, in relative
terms the latter account for almost half (41.3 per cent) of the total subsidies granted
by EU member states; vis-à-vis 39.2 per cent to manufacturing, 8.5 per cent agriculture,
2.5 per cent non-railway transport, 2.4 per cent coal, 1.1 per cent ﬁnancial
services and 5.0 per cent other economic sectors. Moreover, as will be seen below,
some countries inject considerably more public money into their national railway
sectors than others – suggesting a classic question of comparative political economy:
namely, whether the cross-national differences, to some extent, can be explained
by differences in the partisan composition of government.
The remainder of this article is organised as follows. The next section introduces
the data source of the dependent variable. Subsequently, hypotheses are formulated
regarding the issue of how partisan politics may inﬂuence the political allocation
of railway subsidies. There are, for instance, plausible arguments that lead to the
expectation that rail subsidisation is an issue beyond left and right – implying that the
Crössmann and Mause
2 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
use of this speciﬁc policy tool is not inﬂuenced by partisan considerations. The
section after that presents the empirical ﬁndings from a multiple regression analysis.
It turns out that rail subsidisation is not a purely technical issue in the sense of
being completely determined by sector-speciﬁc conditions. Rather, the regression
results suggest that there is room for governments’ partisan preferences. Controlling
for other potential determinants (sector size, public/private ownership, intermodal
competition and so on), government partisanship in fact contributes to the explanation
of the considerable cross-country variation. Contrary to the expectations of traditional
partisan theory, countries governed by left-wing (bourgeois) parties showed
signiﬁcantly lower (higher) subsidy levels. After discussing possible explanations
for this unexpected result, the article concludes with an overview of issues that
demand further research.
Empirical Background: Cross-National Differences in Subsidy Injections
This study is based on a subsidy data set provided by the EC as part of its ‘State Aid
Scoreboard’. The Commission’s data set includes data on ‘subsidies to the railway
sector in million EUR’ for 15 West European countries from 1998 onwards, and
for 10 East European countries that are marked with an asterisk (*) in Table 1. Eight
of the East European countries joined the EU in 2004: Czech Republic (subsidy
data available from 2003 onwards), Estonia (2003–), Hungary (2003–), Lithuania
(2004–), Latvia (2003–), Poland (2003–), Slovenia (2003–) and Slovak Republic
(2004–). In 2007, Bulgaria and Romania became EU member states (for both subsidy
data available from 2006 onwards). The latest version of the Commission’s data
set ends at the end of 2011 (see EC, 2011). As there are many missing values in the
years 2009–2011, our investigation period ends at the end of 2008.
As Table 1 shows, there are considerable differences in subsidy levels between
countries. For a number of different indicators displayed in Columns (1)–(6), we
calculated averages for the whole investigation period 1998–2008. Given the data
limitations mentioned above, for the 10 East European countries these averages could
only be based on a smaller number of years (that is, 3–6 years). In absolute terms
(Column 1), subsidy levels ranged from €6 million in Lithuania, which can be
grouped together with Estonia (€13 million), Latvia (€25 million) and Portugal
(€47 million) into a ‘relatively-low-subsidy’ country cluster, to €4.8 billion in the
United Kingdom, €6.4 billion in Italy, €8.5 billion in France and €9.3 billion in
Germany.
Bivariate correlations between the subsidy levels displayed in Column (1) and
population size as well as geographical size (area in km2
) show that subsidy levels are
signiﬁcantly higher in more populous (ρ = 0.647***) and geographically larger
countries (ρ = 0.392**).1
At the same time, more populous and geographically larger
countries exhibit a signiﬁcantly larger rail network (length of lines in use) as
Rail subsidisation in the European Union
3© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
measured in kms: ρ is 0.842*** and 0.830***, respectively. Moreover the length of
the railway network, in turn, is highly correlated with other indicators capturing
sector size such as, for instance, the number of employees in the railway sector
(ρ = 0.867***) and the number of locomotives (ρ = 0.885***; in both the cases
N = 24 countries as data for United Kingdom is not available). The data on
employees and locomotives is taken from the World Bank’s Railways Database, the
International Union of Railways’ (UIC) Statistics Database and the UNECE
Table 1: Subsidies to the railway sector (EU-25, mean values 1998–2008)
Mio. € € per
capita
Percentage
of GDP
Percentage
of GVA land
transport
Percentage of total
central government
expenditure
In €1000
per line km
(1) (2) (3) (4) (5) (6)
Luxembourg 292 642 1.17 38.49 3.80 1064
The Netherlands 2446 152 0.56 24.77 1.94 868
Belgium 2449 236 0.92 38.89 3.00 702
Italy 6425 111 0.52 16.44 1.77 392
United Kingdom 4798 80 0.27 15.07 0.71 289
Denmark 780 145 0.43 18.83 1.16 288
France 8457 140 0.56 25.67 2.25 286
Germany 9332 113 0.44 28.12 3.14 263
Ireland 481 118 0.38 30.48 1.19 250
Greece 446 41 0.29 22.05 0.81 184
Slovenia* 187 93 0.71 24.44 2.19 152
Austria 750 92 0.34 11.95 1.15 131
Sweden 1095 122 0.38 15.77 1.21 99
Finland 458 88 0.32 11.84 1.21 78
Spain 1138 27 0.17 6.57 0.87 77
Hungary* 576 57 0.92 21.39 1.92 73
Slovakia* 232 43 0.75 14.55 2.41 64
Czech Republic* 289 28 0.36 6.65 0.83 30
Bulgaria* 95 12 0.45 9.04 1.09 23
Romania* 189 9 0.29 2.43 0.47 18
Portugal 47 5 0.04 1.92 0.09 17
Estonia* 13 10 0.15 3.03 0.42 14
Poland* 231 6 0.10 2.93 0.35 12
Latvia* 25 11 0.18 3.90 0.64 11
Lithuania* 6 2 0.03 0.42 0.11 3
Mean EU-25 1649 95 0.43 15.83 1.39 216
Mean EU-15 West 2626 141 0.45 20.46 1.62 333
Mean EU-10 East 184 27 0.39 8.88 1.04 40
Notes: Own calculations based on EU State Aid Scoreboard data. GVA = gross value added land transport
(Eurostat/OECD data; disaggregated GVA data for rail transport not available). Countries ranked
according to Column (6). Mio. € = in million EUR.
Crössmann and Mause
4 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
Transport Division Database (UNECE stands for United Nations Economic Commission
for Europe). The ﬁgures suggest that larger railway systems tend to be more
costly to operate than smaller ones and, therefore, tend to have a higher demand
for subsidies.
Controlling for sector size in terms of network length by considering subsidy
levels in EUR per line km (see Column 6), we get another country ranking. In
this perspective, Belgium, the Netherlands and Luxembourg are the ‘big spenders’.
Each km of their railway network is publicly subsidised by €702 000, €868 000
and €1 064 000, respectively. By contrast, with the exception of Slovenia (ranked
11 out of 25), East European countries can be found in the lower half of the subsidies
per line km ranking. Statistically, the clear East–West divide is reﬂected in the huge
differences between the mean values of the groups of East and West European countries
displayed at the bottom of Table 1. Moreover, there is a signiﬁcant negative correlation
between the dummy variable EAST (1= East European country) and subsidies per line
km (log): point biserial correlation coefﬁcient (pbis) is −0.706***.
Furthermore, it is noteworthy that, comparing their 2008 values with the countries’
initial values (that is, the ﬁrst year for which subsidy data is available), only two
states experienced subsidy reductions: Greece and Spain reduced subsidies per line
km between 1998 and 2008 by 24.7 per cent and 34.1 per cent, respectively. By
contrast, the other 23 states increased their subsidy expenditures. These increases
were not driven by price increases as the EC’s ﬁgures are inﬂation-adjusted. That
there is no clear downward trend throughout our country sample stands in contrast to
the development of industrial subsidies where a remarkable cutback can be observed
in the EU and OECD world from the early 1980s to the mid-2000s (Obinger and
Zohlnhöfer, 2007; Zahariadis, 2008; Keman, 2010). A simple but plausible explanation
for this interesting difference is that – unlike in the area of industrial subsidies –
the EU State aid regime or other supranational institutions (for example, the World
Trade Organisation) do not exert pressure on our sample countries to reduce subsidy
payments to the railway industry – rather, the opposite is true.
Of course, the EC can prohibit a particular railway State aid. More precisely, the
Commission can declare State aid ‘which distorts or threatens to distort competition
by favouring certain undertakings or the production of certain goods […], in so far as
it affects trade between Member States’ as being ‘incompatible with the internal
market’ (see Article 107 of the Treaty on the Functioning of the EU). Yet, public
subsidies to the railway sector are justiﬁed and regarded as desirable if they promote
the liberalisation of the rail transport sector, the shift of trafﬁc to the (relatively)
climate-friendly way of rail transport (regarded as a ‘clean mode of transport’), the
reduction of noise pollution and the technical harmonisation of European rail
networks (see EC, 2008, for a detailed description of the derogation criteria). Owing
to rail transport’s classiﬁcation as a ‘service of general interest’, railway subsidies are
furthermore justiﬁed ‘… if they represent reimbursement for the discharge of certain
obligations inherent in the concept of a public service’ (Article 93 of the Treaty on
Rail subsidisation in the European Union
5© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
the Functioning of the EU). These payments for so-called ‘public service obligations’
(for example, public support to rail transport in remote regions) do not, however,
have to be reported by the member states to the Commission and are not included
in its State Aid Scoreboard.2
It should now be clear that the EU State aid regime leaves wide scope for national
governments to justify their ﬁnancial support to the railway industry. As all countries
included in our analysis are subject to the sketched State aid rules, in what follows
we do not further discuss this legal factor of inﬂuence (most similar systems design;
Przeworski and Teune, 1970). Instead, we now turn to our main research question,
namely, whether the factor ‘partisan politics’ contributes to the explanation of the
substantial cross-country differences.
Partisan Politics and Rail Subsidisation: Hypotheses
From the perspective of partisan theory in political economy (for surveys, see Hibbs,
1992; Schmidt, 1996, 2010), an obvious question is whether political parties make
a difference regarding the allocation of subsidies to industrial enterprises, railway
undertakings and other sectors of the economy. In the literature on the political
economy of public subsidies it is often hypothesised that left-wing parties are more in
favour of using the interventionist policy instrument of subsidisation than right-wing
parties (Blais, 1986; Zahariadis, 2002; Obinger and Zohlnhöfer, 2007). Left-wing
parties are supposed to consider state interventions (for example, in the form of
subsidies, regulations and public entrepreneurship) as essential for safeguarding the
functioning of the economy and society as a whole. This especially holds true for
areas of the economy regarded as providing ‘public services’ such as, for example,
the railway sector. By contrast, right-wing parties are often assumed to prefer a lower
level of state interference in the economy and to favour market solutions in the ﬁeld
of economic policymaking (Obinger and Zohlnhöfer, 2007, p. 199). If the depicted
ideological differences hold, then the partisan composition of government should
affect the allocation of railway subsidies as formulated in our ﬁrst partisan
hypothesis: Holding other politico-economic factors constant, the level of railway
subsidies is higher (lower) in countries with strong left-wing (right-wing) parties in
government (Hypothesis 1).
Furthermore, it has to be taken into account that since the 1970s Green parties
entered the political systems of several European countries (Müller-Rommel and
Poguntke, 2002). Important policy objectives of Green parties are environmental
protection and the reduction of environmentally harmful emissions (see, for example,
European Green Party, 2006). In this regard, rail transport is a relatively environment
friendly mode of transport. In 2009, 0.6 per cent of total carbon dioxide emissions
from transport in the EU-27 were produced by railways (excluding indirect emissions
from electricity consumption) vis-à-vis 71.7 per cent road transport, 14.6 per cent
Crössmann and Mause
6 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
navigation, 12.3 per cent civil aviation and 0.8 per cent other transport (EC, 2012b,
p. 129). It can be expected, therefore, that Green parties favour subsidies to the
railway sector in order to give a competitive advantage to passenger and freight
transport via rail. Hence, our second partisan hypothesis is as follows: Holding other
politico-economic factors constant, the level of railway subsidies is higher in
countries with strong Green parties in government (Hypothesis 2).
The ﬁrst two hypotheses stress expected differences between different party
families which possibly have an effect on the political allocation of subsidies. Yet,
it might be argued that rail subsidisation is a rather unlikely case for ﬁnding any
partisan effects at all. As explained above, this sector belongs to the so-called
‘services of general interest’ in the EU. It can be expected that many citizens are
(potential) rail passengers who have an interest in a well-functioning railway system
(Perkins, 2005, p. 9). Granting subsidies in the form of state support to infrastructure
projects (for example, new high-speed rail lines, building new and restoring old
railway stations) or the improvement of service quality (new railway lines, new
rolling stock, faster rail connections and so on) may be used by governments of
whatever partisan alignment to demonstrate that they are doing something good for
this speciﬁc sector and the economy as a whole (for example, facilitating mobility,
creating new jobs and promoting economic growth). Therefore, why should
governments of whatever political ‘colour’ – provided that they have the necessary
ﬁnancial resources – not have an interest in meeting their citizens’ demand by
publicly supporting the railway industry? All this leads to our third partisan
hypothesis: Holding other politico-economic factors constant, there is no systematic
relationship between the partisan composition of government and the extent of rail
subsidisation. The conjecture that the public money injected into the rail industry
does not have a partisan colour is the null hypothesis (Hypothesis 0) against which
the partisan-difference Hypotheses 1 and 2 are tested.
Empirical Strategy
To empirically test the partisan hypotheses formulated above, we use two alternative
approaches of operationalising the strength of political parties in government. First,
a set of variables measuring the cabinet seat share (as a percentage of total cabinet
posts) of different ideological ‘families’ of parties is constructed. Like in Armingeon
et al (2012), social-democratic, socialist, (post-)communist and Green parties are
classiﬁed as left-wing parties. Moreover, we take a separate look at the cabinet seat
share of Green parties in order to test Hypothesis 2. Following Obinger and
Zohlnhöfer (2007), liberal, conservative and (non-Christian democratic) centre
parties are coded as right-wing parties; and the latter together with Christian
democratic parties are classiﬁed as ‘bourgeois’ parties.
Rail subsidisation in the European Union
7© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
As a second, alternative indicator, we use a government ideology index developed
by Schmidt (1996) which is also used in Zahariadis (2010) and many other
comparative public policy studies. The Schmidt index runs from 1 to 5, measuring
the strength of political parties in government in a certain country and year on a left–
right scale. The coding is as follows: 1 = hegemony of ‘bourgeois’ parties as deﬁned
above (that is, no cabinet members from social-democratic and other left-wing
parties); 2 = dominance of bourgeois parties (cabinet seat share of left-wing parties
less than 33.3 per cent); 3 = balance of power between left-wing and bourgeois
parties (both camps exhibit a cabinet seat share between 33.3 per cent and 66.6 per
cent); 4 = dominance of social–democratic and other left parties (leftist cabinet seat
share greater than 66.6 per cent but less than 100 per cent); 5 = hegemony of socialdemocratic
and other left parties (leftist cabinet-seat share 100 per cent). The data to
calculate the aforementioned partisan variables was kindly provided by Schmidt
(2013). In this data set, the cabinet seat share of different party families (as
a percentage of total cabinet posts) is reported per year, weighted by the number
of days a national government was in ofﬁce in the respective year.
Table 2 below presents the results of a cross-sectional ordinary least squares (OLS)
regression analysis with the level of subsidies per line km (country averages for
whole investigation period) as the dependent variable (in logarithmic form because of
right-skewed distribution). All regression models exhibit no problematic level of
multicollinearity (see the low variance inﬂation factor (VIF) values). Moreover,
with the exception of Model (4) in Table 3, in all models the Breusch–Pagan/
Cook–Weisberg test indicates no heteroscedasticity so that White heteroscedasticityrobust
standard errors were not adopted. Some may argue that using such crosssectional
research design is outdated. Clearly, one drawback is that it includes only a
relatively small number of observations (N = 25 countries). Consequently, only a
limited number of explanatory variables can be simultaneously regressed in order to
leave a sufﬁcient number of degrees of freedom. However, our research question is to
explain the cross-national differences in subsidy expenditures.
To be sure, switching to a time-series cross-sectional (TSCS) design would
increase the number of observations. Given the limitations in the time-series
dimension of the dependent variable in the case of the East European countries (see
above), we would get an unbalanced panel data set with 217 observations. Note,
however, that complete time-series subsidy data for the period 1998–2008 is only
available for a reduced sample of 15 West European countries (N of observations =
165 = 11 years × 15 countries). Moreover, note that panel-data unit-root tests indicate
that this time series is non-stationary. Consequently, including subsidy levels as
the dependent variable in a TSCS regression would be inappropriate as it may
produce inﬂated t-values, inﬂated R2
and spurious regression results (Kittel and
Winner, 2005; Plümper et al, 2005; Wilson and Butler, 2007). Against this
background and following other recent comparative studies analysing different types
of public spending (for example, Obinger and Zohlnhöfer, 2007; Castles, 2007a;
Crössmann and Mause
8 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
Table 2: Regression results: Determinants of subsidy levels
Short-term partisan inﬂuence Medium-term partisan inﬂuence Long-term partisan inﬂuence
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Eastern Europe −2.374*** −2.350*** −2.325*** −2.228*** −2.184*** −2.447*** −2.349*** −2.222*** −1.975*** −2.480*** −2.327***
[5.51] [5.00] [5.31] [4.87] [5.60] [5.05] [5.57] [4.86] [6.02] [5.56] [6.06]
Road passengers −0.109 −0.088 −0.101 −0.071 −0.122* −0.097 −0.117 −0.072 −0.164** −0.123 −0.158**
[1.43] [1.10] [1.30] [0.90] [1.74] [1.21] [1.54] [0.91] [2.74] [1.55] [2.17]
Bourgeois government 0.017* — — — 0.035** — — — 0.049*** — —
[1.85] [2.79] [4.56]
Leftist government — −0.008 — — — −0.017 — — — −0.025* —
[0.90] [1.13] [1.83]
Government
ideology index
— — −0.376 — — — −0.681* — — — −0.951***
[1.52] [2.03] [2.98]
Green parties — — — −0.009 — — — 0.004 — — —
[0.15] [0.04]
Constant 21.68*** 20.93** 22.82*** 19.04** 21.89*** 22.17*** 25.14*** 19.07** 24.91*** 24.75*** 29.61***
[3.06] [2.74] [3.04] [2.55] [3.38] [2.86] [3.36] [2.55] [4.57] [3.25] [4.11]
Adjusted R2
0.54 0.49 0.52 0.47 0.61 0.50 0.55 0.47 0.73 0.54 0.62
F-statistics 10.59*** 8.71*** 9.81*** 8.15*** 13.75*** 9.06*** 11.09*** 8.13*** 23.15*** 10.56*** 14.54***
VIF-values ⩽1.11 ⩽1.15 ⩽1.09 ⩽1.01 ⩽1.07 ⩽1.28 ⩽1.11 ⩽1.01 ⩽1.16 ⩽1.24 ⩽1.19
N 25 25 25 25 25 25 25 25 25 25 25
Notes: Dependent variable: Railway subsidies per line km (log mean 1998–2008). The measurement of short-term (Models 1–4), mid-term (5–8) and long-term
(9–11) partisan inﬂuence is explained in the text. Unstandardised coefﬁcients of OLS regressions. Absolute value of t-statistics in brackets. Levels of statistical
signiﬁcance: *10 per cent, **5 per cent, ***1 per cent.
RailsubsidisationintheEuropeanUnion
9©2014MacmillanPublishersLtd.1472-4790ComparativeEuropeanPolitics1–22
Table 3: Regression results: Determinants of subsidy levels (alternative speciﬁcations)
(1) (2) (3) (4) (5) (6) (7) (8)
Eastern Europe −1.725*** −1.955*** −1.665*** — −1.828*** −1.716*** −1.630*** −1.735***
[4.14] [5.82] [3.83] [5.64] [4.57] [4.28] [5.01]
Road passengers −0.163** −0.168** −0.154** — −0.145** −0.160** −0.111 −0.165***
[2.72] [2.73] [2.54] [2.48] [2.70] [1.67] [2.86]
Bourgeois government 0.049*** 0.049*** 0.049*** — 0.041*** 0.050*** 0.051*** 0.053***
[4.56] [4.44] [4.61] [3.69] [4.70] [4.90] [4.99]
Road freight 0.011 — — — — — — —
[0.97]
Public ownership — −0.127 — — — — — —
[0.50]
Public debt — — 0.007 — — — — —
[1.08]
Past subsidy level — — — 0.704*** — — — —
[5.38]
BeNeLux dummy — — — — 0.930* — — —
[1.75]
Rural population — — — — — −0.021 — —
[1.33]
Government size — — — — — — 0.056 —
[1.61]
Trade union density — — — — — — — 0.015
[1.66]
Constant 23.93*** 25.59*** 23.38*** 3.77*** 23.31*** 24.96*** 17.16** 24.21***
[4.31] [4.47] [4.17] [2.51] [4.41] [4.66] [2.41] [4.61]
Adjusted R2
0.73 0.72 0.73 0.80 0.75 0.74 0.75 0.75
F-statistics 17.55*** 16.80*** 17.80*** 28.93*** 19.82*** 18.44*** 19.33*** 19.52***
VIF-values ⩽1.66 ⩽1.16 ⩽1.82 ⩽1.00 ⩽1.38 ⩽1.40 ⩽1.87 ⩽1.24
N 25 25 25 25 25 25 25 25
Notes: Dependent variable: Railway subsidies per line km (log mean 1998–2008). Unstandardised coefﬁcients of OLS regressions. Absolute value of t-statistics
in brackets. Owing to heteroscedasticity, Model (4) is based on White heteroscedasticity-robust standard errors. Levels of statistical signiﬁcance: *10 per cent,
CrössmannandMause
10©2014MacmillanPublishersLtd.1472-4790ComparativeEuropeanPolitics1–22
Busemeyer, 2009; Keman, 2010), we are convinced that cross-sectional regressions
are still an appropriate tool for exploring cross-national differences in public
expenditures.
Do Parties Make a Difference? Multiple Regression Results
The regression results in Table 2 indicate that government partisanship does in fact
help to explain the considerable cross-national differences – but in another way than
expected by the partisan difference hypothesis of traditional partisan theory.
Controlling (i) for sector size (captured by the dependent variable), (ii) for differences
between East and West European countries and (iii) for cross-country differences in
the extent of intermodal competition (the results of these and other control variables
are discussed below), the level of subsidisation was higher (lower) in countries
governed by bourgeois (left-wing) parties. This effect is rather weak in the short run;
that is, when the partisan composition of government is considered only in the period
the subsidies were paid (that is, 1998–2008). Although the variable ‘bourgeois
government’ has a positive sign and is statistically signiﬁcant at the 10 per cent level
(see Model 1),3
the variables ‘leftist government’ and Schmidt’s government
ideology index have a negative sign but remain statistically insigniﬁcant (Models 2
and 3). However, if one follows previous research in comparative public policy (for
example, Castles, 2007b; Schmidt, 2010) and takes into account that the ‘partisan
legacy’ may matter in the sense that the size of subsidy expenditures may to a large
extent be set by decisions of governments in the (distant) past, then the partisan
differences become clearer.
The coefﬁcient of the variable ‘bourgeois government’ is positive and statistically
signiﬁcant at the 5 per cent (Model 5) and 1 per cent (Model 9) level, even if these
regressions are jackknifed (that is, each country removed in turn). The coefﬁcient of
‘leftist government’ is consistently negative and gets statistically signiﬁcant at the
10 per cent level in Model (10); but narrowly misses the 10 per cent level in this
speciﬁcation if single countries are removed (P-values between 0.021 and 0.176).
The coefﬁcient of the government ideology index (higher values indicate stronger
left-wing parties in government) is negative and statistically signiﬁcant at the
10 per cent (Model 7) and 1 per cent (Model 11); when jackknifed, the coefﬁcient
remains signiﬁcant at the 5 per cent level in Model (11), and narrowly misses the
10 per cent level in Model (7) if Slovenia (P = 0.106) and Lithuania (0.124) are
removed.4
Models (5)–(11) consider the partisan composition of government in the medium
term and long term to get a measure of the dominant partisan heritage.5
Medium term
means that the average of the respective partisan variable is calculated for the period
1990–2008 in the case of West European countries. In the case of East European
countries, the average for the period between the ﬁrst democratically elected
Rail subsidisation in the European Union
11© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
government since 1990 (as coded in Schmidt, 2013) until 2008 is taken. Long term
means that the average for the period between the ﬁrst democratically elected
government after the end of World War II (that is, 1945) until 2008 is taken.
Following Schmidt (1996), in the case of Greece, Portugal and Spain, the calculation
period starts after their transition from authoritarian to democratic regimes (that
is, 1974, 1976 and 1977, respectively). In the case of the East European countries, the
long-term calculation period starts with the ﬁrst democratically elected government
after the opening of the ‘Iron Curtain’ in 1990.
Interestingly, exploring the determinants of industrial subsidies, Neven (1994) and
Zahariadis (2010) also get the regression result that subsidy levels are lower (higher)
under leftist (bourgeois) governments. One possible explanation for this empirical
ﬁnding might be that leftist and bourgeois parties have different clienteles (that
is, party members, voters and interest groups), which they are trying to satisfy
(see Zahariadis, 2010, p. 445). Given the fact that the railway sector is not only an
important input factor for a well-functioning economy (transportation of goods) but
at the same time a proﬁtable source of revenue for other economic entities (contracts
for business ﬁrms constructing tracks, locomotives, train stations and so on), granting
subsidies to the railway industry may be perceived as one way of doing something
good for the private sector of the economy. As bourgeois parties are often said to be
more ‘business friendly’ than leftist parties, this might explain why they tend to inject
more public money into this industry although their public rhetoric often follows
market-liberal principles.
By contrast, railway workers and their trade unions as core supporters of left
parties may prefer ‘State aid’ directly accruing to them over subsidy payments
directly beneﬁtting railway ﬁrms. To provide help to their core clientele, left-leaning
parties in power can (try to) implement stricter employment protection rules,
generous old-age provision schemes, muniﬁcent unemployment beneﬁts in the case
of layoffs or regulations creating better working conditions. These types of state
support, however, do not (directly) show up in the EU State Aid Scoreboard data.
Finally, regarding the Green party hypothesis, the regression analysis suggests that
Green parties do not make a difference in the ﬁeld under investigation. Holding other
factors constant, subsidy levels were not signiﬁcantly higher in countries with a highcabinet
seat share of Green parties (averages 1998–2008 and 1990–2008; see Models
4 and 8 in Table 2). This also holds for other operationalisations of the Green party
variable (detailed regression results not reported): the percentage of years with Green
party participation in government (both in the short run, 1998–2008, and longer run,
1990–2008); and dummy variables getting 1 if the Greens were part of at least one
government during the respective time periods (that is, 1990–2008, 1998–2008).
A possible explanation for this ﬁnding is that Green party power in national government
was rather low in the considered country sample and time period. According
to Schmidt (2013), Green parties were part of national governments in less than
half of the countries (11 of EU-25): Belgium (1999–2003), Bulgaria (1995/1996),
Crössmann and Mause
12 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
Czech Republic (2007/2008), Finland (1995–2002; 2007/2008), France (1997–
2002), Germany (1998–2005), Ireland (2007/2008), Italy (1992–1994; 1996–2001;
2006–2008), Latvia (2002–2008), Romania (1991/1992) and Slovenia (1991–1993).
Hence, over the period 1990–2008, only in 59 out of 475 country-years (that is, 12.4
per cent) there was Green government participation. Moreover, only in two countries
did the Green cabinet seat share exceed a threshold of 17 per cent: in Germany in 7
years (maximum 21.4 per cent) and in Latvia in 4 years (maximum 27.9 per cent).
Drawing on ‘contagion theory’ (for example, Hicks and Swank, 1992), another
explanation for the fact that Green parties do not matter in the considered context is
that the main parties in the respective political systems might simply have adopted
Green policies such as, for instance, shifting more trafﬁc from roads to railways.
Control Variables and Robustness Checks
It is clear ‘… that the party composition of government is but one variable among
a wide variety of determinants of public policy, and that a ‘partisan theory’ (Hibbs,
1992) of public policy is but one approach to the comparative study of policy outputs
among other theories’ (Schmidt, 1996, p. 169). On the basis of the politico-economic
literature on the determinants of government subsidies in other areas of economic
policy (for example, Blais, 1986; Neven, 1994; Obinger and Zohlnhöfer, 2007;
Zahariadis, 2008, 2010; Keman, 2010), it is hypothesised that the allocation of
subsidies to the railway sector is also inﬂuenced by socio-economic and sectorspeciﬁc
factors (for example, sector size, intermodal competition and public/private
ownership), the public budget situation, and past subsidy commitments (that is,
policy inheritance). In so doing, we analytically consider subsidisation as being
a governmental allocation decision affected by partisan preferences and certain
constraints a government is confronted with.
Socio-economic and sector-speciﬁc conditions
An obvious explanation for the cross-country differences, which has already been
identiﬁed above, is the factor ‘sector size’. A high level of subsidisation may just
reﬂect the high subsidy requirements of a large railway network. In fact, there is
a statistically signiﬁcant positive correlation between rail subsidies and line kms
indicating that countries with a relatively large railway network also show a relatively
high level of rail subsidisation (r = 0.595***). In a bivariate regression with the
absolute level of subsidy expenditures (log mean 1998–2008) as the dependent
variable, network length (log mean 1998–2008) already explains 36 per cent of the
variation between countries (detailed results not reported). Hence, to control for
Rail subsidisation in the European Union
13© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
sector size, we used subsidies per line km as the dependent variable of the multiple
regression analysis.6
As Table 2 shows, the coefﬁcient of the dummy variable ‘Eastern Europe’ is
negative and statistically signiﬁcant in all regression models. This is not surprising.
The results of the bivariate correlation analysis and the comparison of means
presented above have already revealed that subsidy levels are considerably lower in
East European countries. At least three factors help explain this East–West divide:
the degree of economic development, the level of economic growth and the extent
of carbon dioxide emissions (details below). These factors are not only highly
correlated with the East dummy but are also highly correlated with each other. Thus,
to avoid multicollinearity problems, we decided to include the East dummy as
a kind of ‘master control variable’ in all regressions.
First of all, East European member states show a signiﬁcantly lower level of
economic development as measured by real GDP per capita (pbis = –0.869***).
At the same time, the degree of economic development is an important explanation
of the cross-national differences in the extent of rail subsidisation. There is
a statistically signiﬁcant positive correlation between real GDP per capita and
subsidies per line km (ρ = 0.810***). Through the lens of socio-economic theories
of public policy (Peacock and Scott, 2000; Wilensky, 2002), an interpretation for this
ﬁnding is that citizens and companies in richer countries might (i) demand a higher
quality of railway services and infrastructures and/or (ii) be accustomed to a higher
service and infrastructure quality. In fact, there is a statistically signiﬁcant positive
correlation between real GDP per capita and the index variable ‘quality of railroad
infrastructure’ (mean 2001–2008; higher index values indicate higher quality;
ρ = 0.538***), which is one item of the Global Competitiveness Index provided by
the World Economic Forum (2012). In East European countries, which exhibit
a comparatively low level of economic development (see above), rail infrastructure
quality is signiﬁcantly lower than in West European countries (pbis = –0.509***).
A comparatively high societal demand possibly translates into a comparatively
high demand of the national railway industry for public subsidies (for example,
when higher quality implies higher costs). At the same time, governments in more
economically prosperous countries might be able to meet their citizens’ and companies’
demands because of larger tax revenues compared with countries that are
economically less advanced.
Moreover, East European countries have experienced signiﬁcantly higher levels of
economic growth (real GDP growth) compared with their fellow member states
in Western Europe (pbis = 0.658***). At the same time, economic growth is an
important explanation of the cross-national differences in the extent of rail subsidisation.
There is a statistically signiﬁcant negative correlation between real
GDP growth and subsidies per line km (ρ = –0.589***). This suggests that there
is an inverse relationship between economic growth and subsidy expenditures as
hypothesised in studies investigating industrial subsidies (for example, Obinger
Crössmann and Mause
14 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
and Zohlnhöfer, 2007). There is a plausible explanation for the fact that subsidies
to the railway sector are higher in times of economic slump: lower freight volumes
and declining passenger trafﬁc may result in lower revenues from freight and
passenger transport, leading to an increased demand for and supply of public
subsidies.
Furthermore, East European countries exhibit lower levels of carbon dioxide
emissions from transport (in million tonnes; pbis = −0.667***). At the same time,
carbon dioxide emissions are an important explanation of the cross-national differences
in the extent of rail subsidisation. There is a statistically signiﬁcant positive
correlation between carbon dioxide emissions from transport and subsidies per
line km (ρ = 0.545***). A plausible explanation for this positive association is that
policymakers in high-emission countries inject a comparatively high level of public
money into the railway sector in order to try to shift trafﬁc towards this relatively
low-emission transport mode.
Intermodal competition
When it is assumed that many policymakers today have environmental concerns
in mind and use subsidies as a policy instrument to shift trafﬁc from road to rail
transport, then an obvious control variable is the degree of competition between
these modes of transport (so-called intermodal competition). Contrary to expectations,
in all regression models the variable ‘road passengers’ measuring the share of
passenger transport via cars and buses in total inland passenger transport
(percentage of total passenger-kms travelled by trains, cars and buses; Eurostat
data) has a negative sign and reaches statistical signiﬁcance in some models. In
other words, railway subsidy levels, ceteris paribus, are not systematically higher
in countries with high levels of passenger transport via road. A possible explanation
for this unexpected result is that, as noted earlier, payments for so-called ‘public
service obligations’, such as subsidies to promote local passenger trafﬁc by rail, are
not included in the EC’s subsidy data set. Another interpretation could be that
governments in countries in which many citizen-voters prefer to travel on the road
may have little incentive to massively subsidise rail passenger transport as a
competing mode of transport.
As Model (1) in Table 3 shows, the independent variable ‘road freight’, measuring
the share of freight transport via roads (in percentage of total inland freight transport
by railways, roads and waterways in freight tonne kms; Eurostat data), has the
expected positive sign but remains statistically insigniﬁcant. We abstained from
including this variable in all speciﬁcations as it causes a multicollinearity problem:
the share of road freight transport is signiﬁcantly lower in East European countries
(pbis = −0.608***). Also note that we included the modal-split variables in timelagged
form (that is, share of road transport in the year before the ﬁrst year for which
Rail subsidisation in the European Union
15© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
subsidy data is available for a country) because of the danger of reverse causality:
high railway subsidies might cause a higher share of rail transport, while the latter
might cause cuts in subsidy payments.
Public/private ownership
In what follows, we test whether other obvious control variables increase the
explanatory power of the best-ﬁt regression Model (9) in Table 2 that serves as
the baseline model and explains 73 per cent of the cross-country variation of
subsidy levels. It might be argued, for instance, that whether a railway industry is
100 per cent state-owned or (partly) run by private-sector companies makes
a difference in terms of subsidisation. However, as the insigniﬁcant coefﬁcient
of the variable ‘public ownership’ in Model (2) in Table 3 suggests, the degree of
government ownership does not contribute to the explanation of the cross-national
differences in the extent of subsidisation.7
The variable ‘public ownership’ is an
additive index mainly calculated on the basis of an OECD (2011) data set (Conway
and Nicoletti, 2006) that provides information about the ownership status of
‘the largest ﬁrm in operation’ in the sub-sectors rail infrastructure, passenger
transport and freight transport over the period 1975–2007. As we will see below,
in most cases the incumbent state-owned company is still the biggest player. In
addition, information provided in IBM (2011) was taken into account. The coding
in each sub-sector is as follows: 0 = no government ownership; 0.25 = majority
of shares private; 0.75 = majority of shares state owned; 1 = 100 per cent stateowned.
Hence, the highest possible index value for a country in a year is 3. The
variable ‘public ownership’ included in Model (2) in Table 3 represents the mean
value 1998–2008 for each country.
A possible explanation of why ownership status does not help explain the crosscountry
differences could be that the cross-country variance in terms of public/
private ownership is rather low in our country sample. According to the coding and
data sources explained above, there are only four countries that have witnessed major
changes away from public ownership. Denmark and the Netherlands switched to
‘no government ownership’ in the freight sector in 2001 and 2000, respectively. In
Estonia, the incumbent operator Eesti Raudtee, who operates about 70 per cent of the
country’s rail network and performs rail freight transport, was privatised in 2001
(34 per cent of shares still held by the Estonian government) and re-nationalised in
2007 (that is, 100 per cent state-owned). As passenger transport was provided by
private and state-owned operators during our investigation period, this sub-sector is
coded with 0.5 throughout. The United Kingdom switched to ‘no government
ownership’ in the passenger transport sector in 1994, in the freight transport sector
in 1994 and in the infrastructure sector in 1996 – though it should be mentioned
that Network Rail as a private company limited by guarantee, which since 2002 (after
Crössmann and Mause
16 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
the collapse of Railtrack) owns the infrastructure, receives government funding and
also has public members (UK Department of Transport and local authorities) acting
as guarantors (see Newbery, 2006, for more details).8
Fiscal pressure
Furthermore, we found no evidence for the ﬁscal-pressure hypothesis, whereby
subsidy levels should be lower in countries whose leeway for subsidisation is
constrained by a high level of public debt (see, for example, Neven, 1994;
Zahariadis, 2002; Obinger and Zohlnhöfer, 2007; Keman, 2010). Including the level
of general government gross public debt in percentage of GDP (Eurostat data)
yielded statistically insigniﬁcant results for this variable (see Model 3, Table 3).
A statistically insigniﬁcant result also occurs when the level of general government
net debt interest payments in percentage of GDP (Eurostat/OECD data) is taken as
a proxy for ﬁscal stress (detailed regression results not reported). One explanation
for this empirical ﬁnding is that state support for rail transport as a ‘service of general
interest’ might have remained untouched by the pressure for budget consolidation
exerted by the EU’s Stability and Growth Pact regime. As mentioned above, only
2 out of 25 countries – Greece and Spain – in 2008 showed a lower subsidy level than
at the beginning of the observation period. Note that we included the ﬁscal variables
in time-lagged form (that is, public debt and debt interest payments in the year before
the ﬁrst year for which subsidy data is available for a country) because of the obvious
reverse causality (or endogeneity) problem: a high level of subsidy expenditures in
a certain time period (say, 1998–2008) might cause budgetary problems in this
period; and ﬁscal problems in this particular period might cause subsidy-expenditure
cuts in this time period.
Path dependence and policy inheritance
The fact that only 2 out of 25 countries reduced subsidy levels over time illustrates
that ‘path dependence’ (Pierson, 2000) and ‘policy inheritance’ (Rose, 1990) play an
important role in the ﬁeld under investigation. Countries starting on a relatively high
initial level of subsidisation (that is, the ﬁrst year for which subsidy data is available
for a country) also exhibited a relatively high subsidy level in subsequent years.
Statistically, the mean level of subsidy payments (per line km) in the investigation
period is positively and signiﬁcantly related to the extent of subsidies in the past (that
is, the initial level of subsidisation per line km; ρ = 0.949***). The adjusted R2
of
regression Model (4) in Table 3 shows that the initial level of subsidisation explains
80 per cent of the cross-country variation in subsidisation. This result suggests that
it is difﬁcult for a government (regardless of whether leftist or rightist) to massively
Rail subsidisation in the European Union
17© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
cut back the level of subsidies in a country in the short run (say, between 1998 and
2008).
Note that this conclusion is not in conﬂict with the partisan effects reported above:
government partisanship makes a difference in the cross-country perspective,
however, in the within-country perspective – observing the rather short time period
1998–2008 – inheritance seems to be more important than government ideology.
Although the initial level of subsidisation already explains a great deal of the
subsequent cross-country variation, we abstained from including the initial level as
the ‘master variable’ in all model speciﬁcations. Given the short investigation period
and the high stickiness of subsidy spending over this short time period, in our view it
is rather unsatisfactory to ‘explain’ the cross-national differences in subsidy levels
over this period just by the ‘historic’ initial level of subsidisation (that is, the extent
of subsidisation a few years earlier). The latter approach would raise the question of
which factors may explain the cross-national differences in the initial levels.
Finally, it should be noted that, to test whether single countries have a strong
leverage on the ﬁndings, we jackknifed all model speciﬁcations presented in Table 3
by removing each country in turn; the regression results are not sensitive to the
exclusion of any single country. As a further robustness test, we included a BeNeLux
dummy taking on the value of 1 for Belgium, the Netherlands and Luxemburg that
exhibit the highest subsidy levels in our regression analysis and together could be an
‘inﬂuential case’ in statistical terms. Model (5) in Table 3 shows that the inclusion
of the BeNeLux dummy does not signiﬁcantly change the results.
Conclusion
In recent years, a number of empirical studies in the ﬁeld of comparative political
economy have analysed the development and determinants of government subsidies
granted to different sectors of the economy in advanced democracies. The public
subsidisation of railways has been neglected in this literature. This is surprising
given the fact that almost all EU member states – Cyprus and Malta ‘have no railway
operators’ (EC, 2012a) – inject a considerable amount of public money into this
industry year after year. Moreover, the EC’s State Aid Scoreboard reveals that
railway subsidies account for about 40 per cent of the total subsidies granted by
EU member states (2008 ﬁgures).
This article has used a multiple regression framework to investigate what drives
the substantial differences in the extent of rail subsidisation in the EU-25 over
the period 1998–2008. It turns out that subsidy levels are signiﬁcantly higher in
countries with a large railway network and a high degree of economic development.
Controlling for these socio-economic factors and other potential determinants
(public/private ownership, intermodal competition and so on), the regression analysis
moreover suggests that the political allocation of subsidies to the railway industry
Crössmann and Mause
18 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
is not a purely technical issue beyond left and right. By contrast, government
ideology helps to explain the substantial cross-national differences: subsidy levels are
systematically lower (higher) in countries governed by left-wing (bourgeois) parties.
This result is in line with the partisan effects identiﬁed in Neven (1994) and
Zahariadis (2010) for the ﬁeld of industrial subsidies – and is remarkable insofar as
traditional partisan theory within the political economy literature on subsidisation
expects that government ideology works in the opposite direction: that is, left-leaning
governments are expected to be more inclined to ﬁnancially support national
industries.
As the empirical ﬁndings brieﬂy summarised here were already discussed in detail
above, we can ﬁnally point out some promising avenues for further research. It would
be interesting to investigate, for instance, whether the identiﬁed partisan effects
remain robust when the analysed subsidisation period is extended further back in
time. This would, furthermore, allow an examination of the long-term development
of rail subsidisation. Going beyond our macro-quantitative analysis using the EC’s
aggregate-level data for the whole railway industry, it would also be interesting to
explore in more depth exactly how the demand for and supply of subsidies interact.
A deeper analysis of the determinants of rail subsidisation in a comparative
perspective, however, requires more and better (internationally comparable) data
both in the time-series and the intra-sectoral dimension. Which particular railway
enterprises in a country receive public subsidies? For which purposes are the
subsidies paid? Which sub-sectors (network, freight trafﬁc and passenger trafﬁc)
receive railway subsidies? Is the railway sector subsidised via other channels not
disclosed to the Commission (for example, tax exemptions, interest-free loans and
debt guarantees)?
In this context, a special issue worth exploring is whether interest groups play
a role in the ﬁeld of rail subsidisation. In some countries, there might be powerful
railway trade unions and/or associations of railway enterprises lobbying for more
subsidies and against subsidy retrenchment. The coefﬁcient of the variable
‘trade union density’ in regression Model (8) in Table 3, which narrowly misses the
10 per cent signiﬁcance level, points in this direction. However, this is just a proxy
variable measuring union density in the whole economy (that is, union membership as
a proportion of wage and salary earners; mean values 1998–2008; Armingeon
et al, 2012). Hence, a further desideratum for future research is to develop interestgroup
indicators based on comparable cross-country data on the strength and the
lobbying activities of employers’ associations and trade unions in the railway industry.
Acknowledgements
For their helpful comments on earlier drafts of this article, the authors would like to
thank Thorsten Hüller, Henning Schmidtke, Laura Seelkopf, Peter Starke, Stefan
Rail subsidisation in the European Union
19© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
Traub, the three anonymous reviewers of this journal as well as participants of the
2011 Annual Meeting of the European Public Choice Society in Rennes, the 2011
Prague Conference on Political Economy and the 2011 Congress of the International
Institute of Public Finance in Ann Arbor.
About the Authors
Katharina Crössmann is a Research Fellow in the Center for Social Policy
Research (ZeS) and member of the DFG-funded Research Center ‘Transformations
of the State’ (SFB 597), both at the University of Bremen, Germany. Her current
research focuses on the comparative political economy of industrial subsidies.
Karsten Mause is an Assistant Professor of Political Economy in the Institute of
Political Science (IfPol) at the University of Münster, Germany. His current research
interests include the comparative political economy of privatisation, (de)regulation
and subsidisation.
Notes
1 If not otherwise stated, the results of bivariate correlations reported in this article are based on average
values of the respective variables for the period 1998–2008 (N = 25 countries), using Eurostat data.
Depending on the scale level, different correlation coefﬁcients were appropriate and calculated:
r (Pearson’s product-moment correlation coefﬁcient), ρ (Spearman’s rank correlation coefﬁcient), or
(point biserial correlation coefﬁcient) pbis. Where appropriate, the variables were logarithmically
transformed. Levels of statistical signiﬁcance: *10 per cent, **5 per cent, ***1 per cent.
2 The Scoreboard ‘includes all public subsidies communicated to the Commission as well as subsidies that
have been notiﬁed and authorised by the Commission under relevant State aid rules. The ﬁgures exclude
compensation for services of general economic interest’ (EC, 2011).
3 Note, however, that the ‘bourgeois’ coefﬁcient narrowly misses the 10 per cent signiﬁcance level
(P ⩽ 0.10) if Model (1) is jackknifed (that is, each country is removed in turn to test whether the results
are sensitive to the exclusion of single countries): Lithuania (P-value of ‘bourgeois government’ after
removing this country: 0.108), Sweden (0.113), Slovenia (0.142), the Netherlands (0.147), Luxembourg
(0.154) and Portugal (0.163).
4 The coefﬁcients for the right-wing parties (that is, cabinet share of bourgeois parties excluding Christian
Democrats) in the different time periods continuously show a positive sign but remain statistically
insigniﬁcant and are not reported.
5 Schmidt (2010, p. 212) notes: ‘Advanced partisan theory adds to standard partisan theory the distinction
between short-term impacts and the long-term legacies of parties. Partisan effects comprise not only
contemporaneous short-term impacts (measured by a party A’s cabinet seats share at a particular time
point or over a shorter period), but also the long-term legacy of a party A on policy positions in that
society’ (see also Castles, 2007b, pp. 29–30).
6 One might conjecture that subsidies, ceteris paribus, are higher in less-urbanised countries because of
the need to connect rural areas. However, the coefﬁcient of the variable ‘rural population’ (in percentage
Crössmann and Mause
20 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
of total population; World Bank data) included as an additional control remains insigniﬁcant and does
not change the presented regression results (see Model 6, Table 3).
7 As an additional control, we included the general size of the public sector (that is, total general
government expenditure as a percentage of GDP; Eurostat data) as it captures a general propensity to
rely on the public budget in order to ﬁnance public goods. The variable ‘government size’ shows the
expected positive sign but remains insigniﬁcant when added to our baseline model (see Model 7,
Table 3) as it is correlated with the independent variables ‘Eastern Europe’ (pbis = –0.467**) and ‘road
passengers’ (ρ = –0.546***). In any case, the inclusion or exclusion of the size-of-government variable
does not affect our main results.
8 As an alternative speciﬁcation, we included a privatisation dummy getting 1 in the cases of Estonia,
Denmark, the Netherlands and the United Kingdom. Moreover, we tested a UK dummy as the United
Kingdom is the country with the most experience in rail privatisation. In both speciﬁcations, the
respective privatisation variable was statistically insigniﬁcant (detailed regression results not reported)
and did not change the results.
References
AMTRAK. (2008) Public Funding Levels of European Passenger Railroads. Amtrak Ofﬁce of the
Inspector General, Evaluation Report E-08-02, 22 April.
Armingeon, K., Careja, R., Weisstanner, D., Engler, S., Potolidis, P. and Gerber, M. (2012) Comparative
Political Data Set III, 1990–2010 (35 OECD countries and/or EU member countries). University of
Berne, Switzerland, Institute of Political Science.
Blais, A. (1986) The political economy of public subsidies. Comparative Political Studies 19(2): 201–216.
Busemeyer, M.R. (2009) From myth to reality: Globalization and public spending in OECD countries
revisited. European Journal of Political Research 48(4): 455–482.
Castles, F.G. (ed.) (2007a) The Disappearing State? Retrenchment Realities in an Age of Globalisation.
Cheltenham, UK: Edward Elgar.
Castles, F.G. (ed.) (2007b) Testing the retrenchment hypothesis: An aggregate overview. In: The Disappearing
State? Retrenchment Realities in an Age of Globalisation. Cheltenham, UK: Edward Elgar, pp. 19–43.
Conway, P. and Nicoletti, G. (2006) Product Market Regulation in the Non-Manufacturing Sectors of OECD
Countries: Measurement and Highlights. Paris: OECD Economics Department Working Papers no. 530.
European Commission (EC). (2008) Community Guidelines on State Aid for Railway Undertakings.
Ofﬁcial Journal of the European Union 2008/C 184/07, 22 July, pp. 13–31.
European Commission (EC). (2012a) State Aid Scoreboard. Autumn 2012 Update. Brussels, Belgium,
21 December.
European Commission (EC). (2012b) EU Transport in Figures 2012. Luxembourg.
European Commission (EC). (2011) State Aid Scoreboard: Data on State Aid Expenditure (as of
31 December). Brussels, Belgium: European Commission, DG Energy and Transport.
European Green Party. (2006) The charter of the European greens: European green party guiding
principles. Adopted at the 2nd EGP Congress; Geneva, 13–14 October.
Friederiszick, H.W., Röller, L.-H. and Schulz, C.C. (2003) Evaluation of the Effectiveness of State Aid as a
Policy Instrument: The Railway Sector. Berlin, Germany: Humboldt University (mimeo).
Hibbs, D.A. (1992) Partisan theory after ﬁfteen years. European Journal of Political Economy 8(3): 361–373.
Hicks, A.M. and Swank, D. (1992) Politics, institutions, and welfare spending in industrialized
democracies, 1960–1982. American Political Science Review 86(3): 658–674.
IBM. (2011) Rail Liberalisation Index 2011. Market Opening: Comparison of the Rail Markets of the
Member States of the European Union, Switzerland and Norway. Brussels, Belgium: IBM Global
Business Services.
Rail subsidisation in the European Union
21© 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22
Keman, H. (2010) Fading away of state support to the economy? A comparative analysis of the OECD
world (1980–2002). Comparative European Politics 8(3): 281–303.
Kittel, B. and Winner, H. (2005) How reliable is pooled analysis in political economy? The globalizationwelfare
state nexus revisited. European Journal of Political Research 44(1): 269–293.
Müller-Rommel, F. and Poguntke, T. (eds.) (2002) Green Parties in National Governments. London:
Frank Cass Publishers.
Neven, D.J. (1994) The Political Economy of State Aids in the European Community: Some Econometric
Evidence. CEPR Discussion Paper no. 945. London: Centre for Economic Policy Research.
Newbery, D.M. (2006) Privatising network industries. In: M. Köthenbürger, H.-W. Sinn and J. Whalley
(eds.) Privatization Experiences in the European Union. Cambridge, MA: MIT Press, pp. 3–49.
Obinger, H. and Zohlnhöfer, R. (2007) The real race to the bottom: What happened to economic affairs
expenditure after 1980? In: F.G. Castles (ed.) The Disappearing State? Retrenchment Realities in an
Age of Globalisation. Cheltenham, UK: Edward Elgar, pp. 184–214.
OECD. (2011) Product Market Regulation Database, www.oecd.org/economy/pmr.
Otáhal, T. and Pospíšil, T. (2009) Will Czech trains ever reach their destinations efﬁciently? Independent
Review 14(2): 271–287.
Oum, T.H. and Yu, C. (1994) Economic efﬁciency of railways and implications for public policy: A
comparative study of the OECD countries’ railways. Journal of Transport Economics and Policy 28(2):
121–138.
Peacock, A.T. and Scott, A. (2000) The curious attraction of Wagner’s law. Public Choice 102(1/2): 1–17.
Perkins, S. (2005) The Role of Government in European Railway Investment and Funding. Paper
presented at China Railway Investment & Financing Reform Forum, Beijing, September.
Pierson, P. (2000) Increasing returns, path dependence, and the study of politics. American Political
Science Review 94(2): 251–267.
Plümper, T., Troeger, V. and Manow, P. (2005) Panel data analysis in comparative politics: Linking
method to theory. European Journal of Political Research 44(2): 327–354.
Przeworski, A. and Teune, H. (1970) The Logic of Comparative Social Inquiry. New York: John Wiley.
Rose, R. (1990) Inheritance before choice in public policy. Journal of Theoretical Politics 2(3): 263–291.
Schmidt, M.G. (1996) When parties matter. A review of the possibilities and limits of partisan inﬂuence on
public policy. European Journal of Political Research 30(2): 155–183.
Schmidt, M.G. (2010) Parties. In: F.G. Castles, S. Leibfried, J. Lewis, H. Obinger and C. Pierson (eds.) The
Oxford Handbook of the Welfare State. Oxford: Oxford University Press, pp. 211–226.
Schmidt, M.G. (2013) The Partisan Composition of Governments in OECD Democracies, 1945–2012.
Unpublished dataset, University of Heidelberg (Germany), Institute of Political Science.
Wilensky, H.L. (2002) Rich Democracies: Political Economy, Public Policy, and Performance. Berkeley,
CA: University of California Press.
Wilson, S.E. and Butler, D.M. (2007) A lot more to do: The sensitivity of time-series cross-section
analyses to simple alternative speciﬁcations. Political Analysis 15(2): 101–123.
World Economic Forum. (2012) The global competitiveness report. Executive Opinion Survey (different
issues). Geneva, Switzerland: World Economic Forum.
Zahariadis, N. (2002) The political economy of state subsidies in Europe. Policy Studies Journal 30(2):
285–298.
Zahariadis, N. (2008) State Subsidies in the Global Economy. New York: Palgrave Macmillan.
Zahariadis, N. (2010) State aid and partisan government in the European Union. Social Science Quarterly
91(2): 436–454.
Crössmann and Mause
22 © 2014 Macmillan Publishers Ltd. 1472-4790 Comparative European Politics 1–22