European Union budget politics: Explaining stability and change in spending allocations

Introduction

The general budget of the European Union (EU) establishes on a yearly basis how the fiscal resources controlled by the EU are allocated to a range of policies and programmes that are directly financed and generally managed by the European Commission. To date, the EU budget has not attracted much scholarly attention, perhaps because the fiscal resources administered by the EU are only a fraction of the resources controlled by the member states. The vast majority of studies in this field have thus focused on national budgets: on their changing composition, on the factors explaining budget deficits and on the process of reform. There is less theoretical and empirical research on the EU budget, and this has mainly focused on distributional issues and conflicts: how the EU presidency affects the distribution of resources across member states (Aksoy, 2010), the formation of interest coalitions in the adoption of multiannual financial frameworks (MFFs) (Blavoukos and Pagoulatos, 2011) and the distribution of EU funds at sub-national level (Dellmuth and Stoffel, 2012). Other studies have analysed the general pattern of stability and change in budget reform (Citi, 2013) and the likely effects of the Lisbon Treaty on budgetary policymaking (Crombez and Høyland, 2015). However, much less is known about the determinants of budget change, i.e. the factors driving stability and change in the allocation of resources across budget functions.

This article addresses this question and contributes to the existing literature in three ways. First, it analyses whether the re-definition of voting rules and weights in the Council, determined by successive rounds of treaty reforms, have made budget change more (or less) viable. Second, it estimates how the changing ideological composition of the EU's legislative bodies and how a number of other theoretically relevant variables have driven stability and change in the EU budget. Third, by contrasting and comparing the findings with an earlier empirical analysis of the determinants of policy reform in EU regulatory policies (Citi and Justesen, 2014), this article argues that the variables predicting budgetary and regulatory change only partially overlap and that ideological variables play a much more prominent role in budget change than in regulatory policy change.

The empirical analysis is based on a new dataset tracking the evolution of macro areas of expenditure throughout 1979–2013. The original budgetary figures are published on a yearly basis in the official journal of the EU. These figures, however, are affected to a considerable degree by the problem of ‘repackaging’, i.e. the shifting of chapters or other lower-level programmes to new or different titles or the creation of new chapters and titles, which can make data collection and analysis highly unreliable. To tackle this problem, the dataset was developed by applying a specifically devised codebook that maintains constant codes across the whole timeframe, ensuring data consistency and reliability.

In the remainder of the article, I show how partisan politics and the decision-making rules affect EU spending decisions. In particular, I demonstrate that the changing ideological positions of the Council and the Parliament have a significant effect on budget change. This effect, however, is mediated by the EU legislative procedure and can be heavily constrained by it. More specifically, the ability to form winning coalitions in the Council plays an important role in making policy change possible, and becomes fundamental for the adoption of larger-scale policy changes. In addition, the enlargement of the EU to lower-income countries has contributed significantly to the reallocation of EU-level expenditure. As the empirical evidence shows, the accession of the so-called cohesion countries has produced some of the most extensive budgetary changes in the last three decades.

Explaining budget change: Theory and hypotheses

Most of the literature on public budgets has focused on two different but deeply interconnected themes: assessing how budgets evolve in both the short and long run and finding the determinants of stability and change in the composition of public budgets.

The first theme was initially explored by the early incrementalists, who theorised that budgets and public policies in general are reformed in small steps and through a process of ‘muddling through’. This phenomenon of gradual policy change happens – they argued – because policymakers operate under extensive time and information constraints, which impose severe limits on their ability to design large-scale reforms (Davis et al., 1966; Lindblom, 1959; Wildavsky, 1964). Several years later, another group of scholars started with the same set of assumptions to develop a new model of budget and policy reform, which is nowadays widely known as the punctuated-equilibrium (P-E) model (Baumgartner and Jones, 1993, 2002; Jones and Baumgartner, 2005). The central claim of this model is that public policies generally evolve through small-scale policy adjustments. However, the same forces that produce incremental changes (i.e. the bounded rationality of decision-makers, time constraints and information constraints) are also responsible for major departures from the status quo (Jones and Baumgartner, 2005). The mechanism is the following: governments are assumed to have limited time, attention and information. Because of these constraints, they tend to focus their efforts on the few items that are salient on the political agendas, and overlook a number of other socio-economic issues. Yet, if some of the neglected issues reach a threshold of severity that ‘forces’ policymakers to give them attention, the general pattern of incremental adjustments is likely to be interrupted or ‘punctuated’ by a more discontinuous type of change. This model has been successfully tested on a massive dataset tracking the evolution of US federal legislation and the US budget in the long run (Jones and Baumgartner, 2005) and on other large-n cross-national datasets (Baumgartner et al., 2009; Jones et al., 2009). As such, the P-E model is nowadays widely accepted as the state of the art in the analysis of budget stability and change.

The second theme concerns the determinants of budget change. Here the literature is divided between studies focusing on the agenda-setting process, and studies focusing on veto player variables and partisan/ideological variables. The agenda-setting approach, which is the chief component of the P-E model explained above, is a behavioural model of policy choice that is based on the concept of disproportionate information processing (Jones and Baumgartner, 2005: 115–136). Essentially, its central point is that the expansion and contraction of policy issues on the agenda are non-incremental dynamics that predict (or anticipate) policy change. The partisan and veto-player approaches, on the other hand, predict policy stability and change based on the ideological positioning of legislative bodies and on the configuration of formal and de facto veto players participating in the legislative process. For scholars advocating the second approach, agenda setting is not a fundamental factor in explaining policy change, unless agenda dynamics reshape the policy preferences of veto players.

In the specific EU context, predicting budget reform through the lens of the agenda-setting perspective is quite problematic for a number of reasons. One of these is that the EU agenda cannot be tracked easily: the Parliament and the Council, as the two legislative branches, have their own agendas, but neither can initiate legislation. The legislative initiative is left to the Commission, which controls its own agenda and, in part, the global EU agenda. The European Council provides an additional layer of complexity: it has no legislative power at all but defines the general political priorities of the EU, and as such, it contributes substantially to shaping the EU agenda, though in a very indirect way. Moreover, the current scarcity of data on the dynamics of the EU agenda does not allow a full-blown investigation of its impact on policy (and budget) change. ¹ In this article, I will therefore take the second perspective by analysing the impact of partisan and veto player variables on budget change.

Several contributions have relied on veto player theory to analyse the reform of national budgets. Most have examined the impact of partisan politics on the budgets of specific states: Germany (Bawn, 1999; König and Troeger, 2005), Denmark, the United Kingdom and the United States (Breunig, 2006) and on a larger panel of 19 the Organisation for Economic Co-operation and Development (OECD) countries (Bräuninger, 2005). In all these cases, the key dimension being investigated was the distance in policy preferences between the parties of the government coalition or the distance in policy preferences between two successive governments. This distance, which is generally estimated on the basis of party manifesto data or expert survey data, was found to have a significant impact on national budgets in all cases. Nevertheless, we know from veto player theory that the ideological position and distance between veto players are not the only factor affecting policy stability and change. The number of institutional veto players, defined as the constitutional configuration of individual or collective actors whose consent is necessary to change the status quo, is at least as important as the distance in policy preferences (Tsebelis, 2002). Thus, in a larger comparative study, Tsebelis and Chang (2004) estimated the impact of institutional and partisan veto players on the national budgets of 19 OECD countries, finding that both have a significant impact in determining budget stability and change. In this article, I follow this line of investigation to study how these two types of veto player variables drive stability and change in the composition of the EU budget.

The vast majority of empirical analyses on budget reform are focused on the national level. Few studies exist on the reform of the EU-level budget, and these are mainly concerned with distributional conflicts between member states, emphasising either the bargaining behaviour of member states during budget negotiations (Aksoy, 2010; Schneider, 2013) or the patterns of coalition building in the negotiations for the MFF (Blavoukos and Pagoulatos, 2011). Taking a different perspective, Citi (2013) showed that the EU budget is reformed following a pattern that closely approximates the P-E dynamic. However, none of these contributions has identified the factors that explain policy reform and inertia in the EU budget.

Not much is known about the extent to which different configurations of institutional veto players, created or reshaped by successive treaty reforms, have driven the EU budget towards more or less stability. Likewise, it is not entirely clear whether the ideological composition of the EU legislative branches and their distances on the left-right policy space have prompted larger or smaller budget reforms. These questions are particularly relevant, especially in light of the well-known claim that the EU is a political system whose policies are increasingly driven by partisan control in the Council and the Parliament (Hix, 2008: 110–136). If this was the case, we should find that the changing ideological composition of EU legislative bodies affects changes of the EU budget or lack thereof. Moreover, another question that is extremely relevant, both for research and for European citizens in general, is whether the changing political landscape determined by new elections is a sufficient condition to determine a reallocation of expenditure, or whether the EU institutional constraints can severely limit, or even block, budget change driven by partisan politics.

Research design

Measuring budget change

Following an approach developed by Tsebelis and Chang (2004) for measuring budget reform in 19 OECD countries, I conceptualise the EU budget as a vector in an n-dimensional policy space, with each dimension representing the allocation of funds to a specific jurisdiction. Considering that each jurisdiction is allotted a percentage of the general budget, the annual budget can then be expressed as a sequence of percentages (a₁, a_2, …, a _n ). Under these assumptions, it is possible to calculate the ‘extensiveness’ of budget reform by summing the squared deviations from the previous year's appropriations to the same jurisdictions and then taking the square root of the resulting number. The output variable, called budget distance (BD), can be formalised as follows:

BD = ( a 1 , t - a 1 , t - 1 ) 2 + ( a 2 , t - a 2 , t - 1 ) 2 + … + ( a n , t - a n , t - 1 ) 2

(1)

where a_1,t is the percentage of the budget allocated to jurisdiction a₁ at time t, a_1,t-1 is the percentage of the budget allocated to the same jurisdiction the year before, and a _n,t – a_n,t-1 is the difference between the budget allocated to the nth jurisdiction at time t and at time t – 1. In a nutshell, this is a measure of budget volatility that synthesises in a single variable the total amount of yearly changes made to all the jurisdictions of the EU budget. Its general dynamic is represented in Figure 1. The use of percentages instead of real budgetary figures allows me to control for three confounding factors that could artificially inflate the budgetary figures: the changeover from the European currency unit (ECU) to the euro as the official unit of account, the general increase in the amount of budgetary resources determined by the greater number of states in the EU, and the effect of inflation. OPEN IN VIEWER

Figure 1.

The time series of budget distance, by year (1979–2013).

The data were collected through an in-depth analysis of the general budget of the EU from 1979 to 2013. The base year 1979 was chosen because it was the year of the first direct elections to the European Parliament. Given the lack of data on the party affiliation of members of Parliament (MEPs) before this date, it would have been virtually impossible to estimate the Parliament's ideological positioning before that year, as well as its impact on the EU budget.

To make the data collection and analysis more manageable and reliable, I developed a codebook that classifies the entire set of budgetary items under a specific set of macro-categories of expenditure. These are represented in the first column of Table 1. Each macro-category is then further divided into meso-level categories of expenditure (in the second column of Table 1), which are the codes that were materially used in the classification of the entire set of budgetary items. The reason for relying on this specific codebook is that the EU budget contains an enormous variety of micro-categories of expenditure, packaged under a number of EU-defined ‘titles’ and ‘chapters’. These titles and chapters could not be used as actual codes because of their instability: over the years, specific chapters have appeared, disappeared or moved across titles, while some have been relabelled with a different name. Thus, relying on EU-defined titles and chapters would have created a massive measurement error. The macro-categories of expenditure defined in Table 1, on the other hand, are categories that remain constant across the whole timeframe analysed. More to the point, the use of strictly defined and time-consistent macro-categories of expenditure prevents the coder from coding the frequent repackaging of expenditure as budget changes.

OPEN IN VIEWER

Table 1.

Codes used for coding the EU's macro areas of expenditure.

1. Common agriculture and fisheries policy	1.1 European Agricultural Guidance and Guarantee Funds 1.2 Rural development programmes 1.3 Common Fisheries Policy
2. Regional Policy and cohesion	2.1 European regional development funds 2.2 Cohesion funds
3. External relations	3.1 External relations 3.2 Trade 3.3 Enlargement and neighbourhood policy 3.4 Humanitarian aid
4. Social policy and social affairs	4.1 European Social Fund 4.2 Education and youth policy 4.3 Other social policies
5. Research and Technology development	5.1 Framework programme 5.2 Information society and media
6. Administration and other	6.1 Administrative expenditure of Parliament, Council, ECJ, and all the other administrative bodies (except Commission) 6.2 All other expenditure

There is another important reason for using this codebook: over the last 30 years, the EU's competences evolved considerably as it acquired new roles and developed new programmes almost every year. To capture the evolution of expenditure in a time-consistent fashion across this long timeframe, the categories in the codebook are fully backwards compatible, i.e. they are as applicable to the earlier years of the dataset as to the most recent years. The macro-categories of expenditure shown in Table 1 are exactly designed for this type of analysis. Finally, the macro-categories allow us to focus on the general trends of stability and change in expenditure policies, disregarding the ‘noise’ generated by a number of smaller-scale programmes that the EU created over the years. ³

Estimation and results

Before testing the models through ordinary least squares (OLS) regressions, the time series properties of the data were carefully examined. In particular, to check for stationarity of the time series and hence its suitability for OLS regression, I run the augmented Dickey–Fuller (ADF) test for a unit root in the variable BD. The test rejected, by an ample margin, the null hypothesis for a unit root in the series. Moreover, the residuals of all the regressions were inspected for serial correlation by running the portmanteau (Q) statistic for all the models. In all cases, the test failed to reject the null hypothesis of no serial correlation.

Table 2 shows the results of the OLS regressions for the six models. Model 1 tests the impact of the Council's winning coalitions when controlling for the ideological position of the Council, the Parliament and the Commission. The effect of the first variable is positive and highly significant, implying that an improved ability to form winning coalitions in the Council is likely to determine more extensive budgetary changes. The ideological position of the three institutions is measured on a continuous scale, ranging from highly negative scores, denoting extreme left positions, to highly positive scores, representing extreme right positions. A score of zero or close to zero indicates a centrist position. The scores for the Parliament and the Council are weighted averages, where the weights are determined by the size of the parliamentary groups in the case of the Parliament and the number of votes available to a country in the case of the Council.

OPEN IN VIEWER

Table 2.

Explaining budget change in the EU (1979–2013).

Variables	(1)	(2)	(3)	(4)	(5)	(6)
Council winning coalitions, t − 1	0.508***	0.544***	0.526***	0.447***	0.508***	0.687*
	(0.114)	(0.122)	(0.127)	(0.107)	(0.116)	(0.391)
Council left–right position, t − 1	−0.417***	−0.401***	−0.337*	−0.377***	−0.417***	−0.344**
	(0.113)	(0.115)	(0.197)	(0.104)	(0.115)	(0.153)
Parliament left–right position, t − 1	−0.348***	−0.342***	−0.322*	−0.401***	−0.347***	−0.450
	(0.120)	(0.122)	(0.168)	(0.112)	(0.123)	(0.319)
Commission left–right position, t − 1	0.106	0.152	−0.033	0.117	0.104	0.196
	(0.086)	(0.092)	(0.266)	(0.079)	(0.088)	(0.137)
Max left-right distance, t − 1	–	–	−0.010	–	–	–
			(0.013)
Max Pro-Anti EU distance, t − 1	–	–	−0.120	–	–	–
			(0.196)
Council Pro-Anti EU position, t − 1	–	−0.754	–	–	–	–
		(0.480)
Accession of cohesion countries	–	–	–	2.470**	–	–
				(0.950)
Political budget cycle	–	–	–	–	0.165	–
					(0.800)
Financial framework I (1988–1992)	–	–	–	–	–	−1.167
						(1.800)
Financial framework II (1993–1999)	–	–	–	–	–	−0.703
						(1.994)
Financial framework III (2000–2006)	–	–	–	–	–	0.764
						(4.477)
Financial framework IV (2007–2013)	–	–	–	–	–	1.846
						(5.107)
Constant	0.600	1.915	0.864	0.769	0.556	−0.885
	(0.803)	(1.225)	(1.072)	(0.736)	(0.845)	(5.420)
Observations	34	32	32	34	34	34
R-squared	0.657	0.669	0.651	0.723	0.657	0.682

As the regression table shows, if we control for the effect of the winning coalition variable, a one-point increase in the left–right scale (that is, a one-point movement towards the right) produces a highly significant negative score on the output variable. This means that a shift to the right in the ideological composition of the two legislative bodies determines budgetary choices that are more oriented to the status quo. This finding is particularly interesting in light of some earlier research on regulatory policy change in the EU, which found that the same partisan variables had no significant impact on regulatory expansion and dismantling (Citi and Justesen, 2014). The implication is that the factors at play in budgetary and regulatory change only partially overlap and that partisanship plays a more important role in budget reforms than in regulatory reform.

Model 2 includes the same variables as model 1, but in addition tests the impact of the Council's position on the pro-anti EU dimension. As shown in Table 2, the effect is not significant. This finding, however, is not particularly surprising: although it is well known that the pro-anti EU dimension gains salience during negotiations over the general size of the EU budget (i.e. when the member states have to decide whether to pay more or less towards the supranational budget), the same dimension tends to lose relevance in the reallocation of budgetary resources across expenditure areas. In other words, the correlation between member states' position on the pro-anti EU scale and their preferences for reallocation of spending is very low.

Model 3 focuses on the distance of policy preferences between the Council and the Parliament, both on the left–right dimension and on the pro-anti EU dimension. The effects of the two variables are not significant, and this is a finding that might require some interpretation. Leaving aside the pro-anti EU variable, which exhibits very little variation to yield a significant effect (see also Warntjen et al., 2008), the non-significance of the ideological distance might be explained in light of the Council's range of policy preferences, which are likely to be similar to the preferences expressed by the main political parties of the Parliament. This intuition is supported by the findings of a very recent study on the EU budget process after the Lisbon Treaty (Crombez and Høyland, 2015), which demonstrated that the budgetary preferences of member states commanding a qualified majority in the Council overlap extensively with the preferences expressed by the three main political groups in the Parliament.

Model 4 estimates the impact of the enlargement of the EU to include the cohesion countries. The literature has presented considerable qualitative evidence showing that the accession of these countries has historically sparked extensive budgetary bargains, which in turn have resulted in the re-balancing of expenditure priorities (see, for instance, Blavoukos and Pagoulatos, 2011). However, model 4 tests on a larger-n dataset whether there is a systematic relationship between the two variables while controlling for the ideological and veto-player variables. The coefficient of model 4 shows that this is indeed the case: the accession of these countries led to a statistically significant change in the allocation of resources across the main areas of expenditure. Looking more closely at the raw data, one can observe a substantial movement of resources from areas such as the common agricultural policy to regional policy and the European social fund corresponding with these enlargement rounds.

Model 5 controls for the possible impact of an EU-level political budget cycle. There is no specific theoretical and empirical literature on this aspect of the EU political system, and it is not entirely clear whether EU budget appropriations are affected by the prospect of upcoming EU-level elections. Yet there is an extensive body of comparative literature on this phenomenon at the country level (Alt and Lassen, 2006; Goldbach and Fahrholz, 2011; Mink and de Haan, 2006; Shi and Svensson, 2006), and this is indeed a possibility even at the EU level, considering that MEPs control a sizeable amount of fiscal resources and that these can be strategically deployed in ways that increase the MEPs' chances of being re-elected. In this model, I therefore control for the reallocation of expenditures in comparison with the year preceding an EU election year. The regression estimates, however, report a non-significant coefficient for this model, implying that there is no clear evidence of such a cycle at the EU level. This result is coherent with some recent findings in the literature, which have shown that political budget cycles are likely to occur in cohesive, single-party political systems, not in policies characterised by a higher number of veto players (Bojar, 2014).

Finally, model 6 tests the impact of MFFs on the stability and change of the EU budget. MFFs are inter-institutional agreements between the Parliament and the Council that set the maximum level of aggregate expenditure for the following seven years. For this reason, they can be a constraining factor on the development of annual budgets. Hence, their effect is tested by a set of dummy variables in the regression (one for each MFF), while controlling for the effect of winning coalitions and the ideological position of the Council, Parliament and Commission. The coefficients of model 6 show that MFFs have no significant impact on yearly budgets. Interestingly, however, the effect representing the Council's ideological composition remains significant in this model, while the corresponding Parliament variable becomes insignificant, as a substantial part of the variability is now captured by the MFF dummy variables. Theoretically, this can be interpreted by taking into account that it is exactly during the MFF negotiations that the Parliament tends to act as a non-partisan and unitary veto player, typically deploying its full bargaining power to gain larger expenditures in at least some key areas of the budget (social policy, regional policy and research and technology policies are typically the most favoured areas).

The main findings of the analysis are visually represented in Figure 2. The chart compares the coefficients of model 2 and model 4, and shows how model 4 and in particular the variable ‘accession of cohesion countries’ helps to improve the second baseline model. In both models, the winning coalition variable and the ideological position of the Council and the Parliament on the left–right dimension are highly significant. However, model 4 clearly shows that the inclusion of new member states with a lower per-capita income has a strong and highly significant impact on EU budget allocations. The implication is that the process of enlargement to countries with a substantially lower level of economic performance has a direct impact on the EU budget, producing an effect that remains significant even when we control for the Council's voting rules and for the ideological composition of the two co-legislators. OPEN IN VIEWER

Conclusion

This article has analysed how institutional veto players and partisan politics drive stability and change in the EU budget. The analysis, based on an original dataset tracking the evolution of all the macro areas of expenditure in the long run, found evidence that different types of variables drive the EU budget dynamics. Among the veto player variables, the percentage of winning coalitions (a measure of member states' ability to form coalitions in the Council) was found to affect budget change significantly: the higher the percentage of winning coalitions, the more extensive the budget changes. This finding, which confirms what was hypothesised, implies that voting rules in the Council, as well as the distribution of voting weights to member states, plays a fundamental role in determining the viability of budget changes and, in general, the viability of policy reforms in the EU.

Among the variables tracking the political orientation of policymakers, the Commission's ideological position was not found to have an effect on budget change, whereas the Parliament's and the Council's positions were found to have a significant impact on the extensiveness of budget reform when the institutional variables were controlled for. More specifically, a positive one-point movement on the left–right scale (i.e. a movement towards the right of the two legislative bodies) produced a significantly smaller budget change, whereas a one-point movement towards the left determined a significantly wider budget change. This finding is particularly interesting if we consider that a previous large-n analysis of regulatory reform in the EU found no evidence that the same partisan variables have a significant influence on the dynamics of regulatory expansion and dismantling (Citi and Justesen, 2014). We can, therefore, conclude that the variables at play in budgetary and regulatory reform in the EU only partially overlap and that partisan variables play a much more significant role in budget reform than in regulatory reform. A second dimension that is generally considered salient in the EU policymaking process, i.e. the pro-anti EU dimension, was not found to affect the budget significantly.

Another important component of this analysis is the role of policy distance between the Parliament and the Council, with regard to both the left–right dimension and to the pro-anti EU dimension. The empirical analysis did not find any significant effect of these two variables when their ideological and institutional variables were controlled for. This result can be interpreted in light of veto player theory as a lack of sufficient heterogeneity in policy preferences between the Council and the Parliament with regard to the (re)allocation of budgetary resources. To be sure, we know that the policy distance between the Parliament and the Council tends to reach its apex in negotiations concerning the overall size of the EU budget. However, during negotiations on the allocation of resources to specific categories of expenditure, there seems to be much less heterogeneity of preferences.

Finally, an exogenous factor, the enlargement of the EU to the cohesion countries, was found to be responsible for some of the most extensive budgetary reallocations in the last three decades. Other possibly relevant factors, such as an EU-level political budget cycle and the big negotiation rounds leading to the adoption of the MFFs, were not found to have a significant influence once the aforementioned institutional and ideological variables were controlled for.

The main finding of this article is that partisan politics matters in EU budget policymaking in a way that is much more significant and consistent than in EU regulatory policymaking. The influence of partisan politics, however, is mediated by the institutional setting and can be heavily constrained by it. The ability to form winning coalitions, in particular, plays an important role in making policy change possible, and it is fundamental for the adoption of larger-scale policy changes.