Key drivers of EU budget allocation: Does power matter?
aa r X i v : . [ q -f i n . E C ] D ec Key drivers of EU budget allocation: Does powermatter?
Vera Zaporozhets ∗ Mar´ıa Garc´ıa-Vali˜nas † Sascha Kurz ‡ Abstract
We examine the determinants of the EU budget expenditures allocationamong different countries. In line with earlier literature, we consider two alter-native explanations for the EU budget distribution: political power vs. “needsview.” Extending the original data set from Kauppi and Widgr´en (2004), weanalyze the robustness of their predictions when applying a different measureof power and more sophisticated econometric techniques. We conclude thatthe nucleolus is a good alternative to the Shapley-Shubik index in distributivesituations such as the case of EU budget allocation. Our results also show thatwhen explaining budget shares, the relative weight of political power based onthe nucleolus is lower than the predictions of previous studies based on theShapley-Shubik index.Keywords: EU policies, budget allocation, political power, nucleolus, Shapley-Shubik index.JEL codes: D72, D78, H61, O52 ∗ Corresponding Author. Toulouse School of Economics (LERNA, INRA), e-mail: [email protected] † University of Oviedo ‡ University of Bayreuth Introduction
In 2013, the European Union (EU) expenditure budget was around 149 billion,with cohesion, and agricultural and environmental resources being the primary EUpolicies, with shares of 46.8% and 39.8% respectively. Due to the magnitude of thesefigures, the distribution of the EU budget among different countries is a crucial issueto analyze. In particular, we focus on the relative weights of different factors whenexplaining the budget shares corresponding to each EU member.Previous literature (Courchene et al. , 1993; Anderson and Tyers, 1995; Tangermann,1997; Kandogan, 2000; Kauppi and Widgr´en, 2004, 2007) has tested two alternativeexplanations of the EU budget distribution across the members states. The first is a“needs view,” which states that the budget allocation is determined by the principlesof solidarity. According to this hypothesis, the countries with a high agriculture sectorweighting and/or a relatively worse economic situation emerge as the major recipi-ents of the EU budget. In fact, some of the previous studies have focused exclusivelyon this explanation (Courchene et al. , 1993; Anderson and Tyers, 1995; Tangermann,1997). The second explanation is that budget allocation across the members is re-flected by the distribution of their political power. Thus, those countries with morepower in the allocation process could receive larger shares of the budget.Some studies combine both: needs and the power views (Kandogan, 2000; Kauppi and Widgr´en,2004, 2007; Aksoy, 2010). Thus, empirical analysis by Kauppi and Widgr´en (2004)shows the strong prevalence of political power motives.Their results indicate that political power has much higher weight than needs whendetermining the allocation of budget expenditures among EU member states.The overall purpose of this paper is to reconsider the analysis of Kauppi and Widgr´en(2004) and to challenge their conclusions. To do so, we extend the original data set(1976-2001) up to 2012 and introduce alternative model specifications. In contrast tothe existing studies which have used the Shapley-Shubik index as a measure of polit-ical power, we employ a different measure; the nucleolus. It has been argued that the2ucleolus is an appropriate power measure in distributive situations as well as a goodalternative to traditional measures, such as the Shapley-Shubik index. Moreover, weapply sophisticated econometric techniques, which are more suitable for the analysis.The paper is structured as follows. Section 2 provides a brief introduction on theEU budget, including the processes of designing and allocating EU expenditures andrevenues. In Section 3, we discuss the theoretical properties of two different powerindices. Specifically, we highlight the advantages of the nucleolus over other indicesin distributive situations, such as the EU budget allocation. Finally, in Section 4, wespecify a simple empirical model in order to determine the key drivers for EU budgetallocation. Section 5 concludes with a summary of the main findings in addition tosome policy implications.
As mentioned in the Introduction, the EU expenditure budget represents a significantamount of resources. In 2013, total expenditures were 148,468 million. Although thisis not a substantial amount in relative terms (just 1.13% of the EU-27 Gross NationalIncome, GNI), some crucial policies were developed using EU funding. Examples arethe Common Agricultural Policy (now part of a more extensive section on the preser-vation and management of natural resources) or several policies oriented towards theeconomic development of some target regions (cohesion and competitiveness policies).Each EU member also has to contribute to the EU budget, by means of GNI-basedresources (74.3%), VAT-based own resources (9.5%), and traditional own resources(TOR, 10.4%).In 1976, the EU expenditure budget amounted to 7,563 million. In the last decades,the EU budget has been growing. This increasing path can be interpreted as a snap For example, Montero (2005), Montero (2013) and Le Breton et al. (2012) among others. Figures on the EU budget are available at http://ec.europa.eu/budget/figures/index n.cfm. Successive EU reforms have changed the structure of the budget. In this respectthere are some facts that are worth mentioning. At the Brussels European Councilin February 1988, a political agreement on doubling, in real terms, the budget of theStructural Funds between 1987 and 1993 was reached. Subsequently, Member Statesagreed at the Edinburgh European Council in December 1992 that the budget forstructural operations would be further increased, specifically for the cohesion countries(Greece, Ireland, Portugal and Spain). Also in Edinburgh, Member States decidedto strengthen some particular policies, such as research and development, externalactions and financial aid to Central and Eastern European countries. Although therewere several agreements on setting budgetary limits to the growth rate of expenditure,the basis of a stringent budgetary discipline was established in the Agenda 2000agreements. These reforms have all had an impact on the level and the structure ofbudget expenditures, and have led to some changes in the accounting system. Thebudget has therefore undergone some structural reforms, the most significant beingthose of 1992 and 2006.Regarding the procedure for elaboration and approval of the EU budget, thereare several institutions involved. The European Commission, the Council and theParliament participate in the process of elaborating the EU budget. However, over thepast decades, the role of each institution, as well as the voting rules, have undergone The new members were the Czech Republic, Cyprus, Slovak Republic, Slovenia, Estonia, Hun-gary, Latvia, Lithuania, Malta and Poland. A second reading by the Parliament and the definitive adoption marksthe end of the process.In this paper, we consider the voting decisions of the Council (Bindseil and Hantke,1997; Kauppi and Widgr´en, 2004). Although the Parliament has recently increasedtheir weight in the EU decision process, some EU institutional features have sup- Usually, at least a qualified majority is required to adopt budgetary decisions at the Council. The Treaty of Lisbon extended the role of the Parliament. It was signed by the EU memberstates on 13 December 2007 and entered into force on 1 December 2009. From that moment,European Parliament could decide on both compulsory and non-compulsory expenses, extending itspower and responsibilities with regard to the budget making process. With the entry into force ofthe Treaty of Lisbon a new system known as a “double majority” was introduced. It entered into In recent decades, there has been a growing literature, both theoretical and applied, onpower measures. However, as yet, there is no consensus as to the best way to measurepower. While analyzing the distribution of the EU budget among different countries,previous studies have applied the Shapley-Shubik index (SSI) (Kauppi and Widgr´en,2004), one of the mostly commonly used power measures in this context. By contrast,in this study we propose an alternative measure, the nucleolus (Schmeidler, 1969). Inthe following, we provide strong arguments to support our choice. In the subsequentsection, we compare how the two indices perform in practice, and analyze whether theconclusions reached by Kauppi and Widgr´en (2004) are robust with regard to theirchoice of power index.The general discussion on which power measure is best, and which properties itshould possess, remains open. Napel and Widgr´en (2004) therefore divide existing force on 1 November 2014. The Nice system remained applicable during the transition period up to31 October 2014. For an interesting discussion on the Treaty of Nice, see Heinemann (2003). Among others, the approval procedure applied during the period analyzed in this research (1976-2012) and the qualified majority required at the Council to approve the final EU budget. Thus, Felsenthal and Machover (1998) argue that the SSI is a measure of “P-power,” whereP stands for “purse,” and it evaluates a voter’s expected relative share of a fixed budget. As weargue in the paper (see also Le Breton et al. (2012)) the nucleolus can also be considered a powermeasure in the distributive setting, and can be a good alternative to the SSI. Consequently, in thispaper we focus on these two measures. At the same time, the SSI can also be considered a measureof “I-power” (Felsenthal and Machover, 1998), which assesses the probability of a player casting adecisive vote. Other measures of I-power include the Banzhaf index, the Johnston index, and theDeegan-Packel index. According to the second requirement, one would not want the power analysis tobe extremely sensitive to the details of the game form used to describe the non-cooperative decision process. In the following, we show that only the nucleolus passes For instance, Gul (1989) constructs a non-cooperative game mimicking bargaining process inthe markets. One of the main results state that the payoffs associated with efficient equilibriaconverge to the agents’ Shapley values as the time between periods of the dynamic game goesto zero. Even though Gul’s bargaining procedure is very natural, his results are not relevant formajority games. Some examples of a less natural bargaining procedure but more general resultsare Hart and Mas-Colell (1996) and Vidal-Puga (2008). As for the nucleolus, it has been provedto correspond to the vector of expected payoffs in the legislative bargaining game with randomproposers according to Baron and Ferejohn (1989), where voters directly put forward proposals andvote over the division of a budget. If proposal probabilities coincide with the nucleolus, then thenucleolus is the unique vector of expected payoffs (Montero, 2006). The equality of the expectedpayoffs to the nucleolus also holds for other proposal probabilities depending on the voting game. et al. , 2013). The idea behind the bargaining set is that when theplayers decide how to divide the worth of the coalition, the player who is not satis-fied with the proposed share may object to it. The objection goes against anotherplayer, calling for this player to share their part with the objecting one. The playeragainst whom the objection is made may (or may not) have a counter objection. Anobjection which does not have a counter objection is called justified. The bargainingset consists of all imputations in which no player has a justified objection against anyother player.It seems that the bargaining set properly describes the decision-making procedurewithin EU institutions (see Section 2). Additionally, one of the properties of thebargaining set is that, contrary to the core, it is never empty. However, the bargainingset is often large, in which case there is the problem of choosing a unique outcome. Insuch cases, the nucleolus is a good candidate, since it always exists, it is unique and itbelongs to the bargaining set. On the contrary, in general the Shapley value is not inthe bargaining set. The following example supports this argument. Let us considerthree individuals with individual one being a vetoer. This means that a decision ispassed only when player one is present in a group voting for the decision, however ifthey are on their own, they cannot get the decision passed. In such a situation, thecore, the nucleolus, and the bargaining set coincide, and attribute the whole surplusto player one. On the contrary, the Shapley-Shubik index is (2 / , / , / In this subsection we provide computations of the SSI and the nucleolus for the firstEU Council of Ministers (1958 - 1972). During that period the Council consistedof representatives from six countries. The three “big” countries (Germany, Italyand France) held four votes each, the two “medium” countries (Belgium and theNetherlands) held two votes each and the “little” country (Luxembourg) held onevote. A qualified majority was set at 12 out of 17, i.e., passing a decision requiredat least 12 votes in favour of the decision. As has been highlighted in a number ofstudies, Luxembourg was powerless in such a situation. Since other member statesheld an even number of votes, Luxembourg was never formally able to make anydifference in the voting process. The results are summarized in Table 1.According to the nucleolus, a “medium” country receives half as much weight as a“big” country. This is quite intuitive, since in a minimal winning coalition, a “big” For example, Felsenthal and Machover (1997), among others. A minimal winning coalition is a group of countries whereby if they all vote in favour of a decision,it is passed. Furthermore, none of the countries can be excluded, i.e., if one of the countries changethe vote from “yes” to “no” the decision can no longer be passed. In this scenario, there are twotypes of the minimal winning coalitions: three ”big” countries, or two “big” countries and two“medium” countries.
The Council of Ministers (1958 - 1972).Country Weight SSI Nucl
Germany 4 0 .
233 0 . .
233 0 . .
233 0 . .
150 0 . .
150 0 . Quota
Quota (%) . Asa consequence, in this case the nucleolus treats all minimal winning coalitions equally.It prescribes the total “wealth” for both types of coalitions as being equal to 0 . .
766 and 0 . As a result, the nucleolus is very different from the SSI and other indices in thisexample. For a more detailed discussion see, for example, Montero (2005). See, for example, Montero (2005). Empirical Application
As explained in the Introduction, this paper aims to identify significant key driversand trends for EU budget allocation. In order to discuss the findings of Kauppi and Widgr´en(2004), we extend their data set to include observations for the period 1976-2012. In-terestingly, this period covers different phases of EU integration: from 1976 to 1980(EU9), from 1981 to 1985 (EU10), from 1986 to 1994 (EU12), from 1995 to 2003(EU15), from 2004 to 2006 (EU25) and from 2007 to 2012 (EU27). In this respect, a general model will be proposed, where the budget share on thewhole EU budget of each country depends on an index of political power and a setof variables representative of budgetary needs. The empirical model is presented asfollows: b it = f ( p it , Z it ) + u it , (1)where b it is the percentage of the total EU expenditure budget allocated to country i in the year t , p it is an index of political power for country i and period t , and Z it is avector of factors representative of country i ’s needs in period t . Finally, u it representsthe error term.We have proposed two alternative specifications of the dependent variable, follow-ing the procedure suggested by Kauppi and Widgr´en (2004). On the one hand, weconsider the total expenditure budget share that each country gets in the negotiationprocess ( exp ). On the other hand, an alternative variable is defined, introducing anadjustment to take into account the UK’s budget rebate and other similar compen- Although data for 2013 are available, this last year was removed from the dataset. This is thefirst year with EU-28 with Croatia as a new EU member, so in the context of unbalanced paneldata methods, this observation would need to be dropped since there is only one period for thatobservation (Bluhm, 2013; Wooldridge, 2010 a ). exp adj ). We also use some of the original variables proposed by Kauppi and Widgr´en (2004)as independent variables. First, two different alternatives to measure political powerdiscussed in the previous section are included in the analysis; namely, the SSI ( p ssi )and the nucleolus ( p nucl ). The latter power index was not originally included byKauppi and Widgr´en (2004), but has been considered for a comparison to be madewith the SSI. Additionally, needs are shown using a set of variables ( Z ): each country’sshare of the total agricultural production ( agri ), and the ratio of each country’s GDPper capita and the EU wide GDP per capita ( income ). Table 2 shows some descriptivestatistics of the main variables: This rebate was a compensatory payment made to the UK government in 1985. The mainargument in the rebate negotiations was that a high proportion of the EU budget was spent onthe Common Agricultural Policy (or CAP), which benefits the UK much less than other countries,as it has a relatively small farming sector as a percentage of GDP. The compensation consists ofreallocating some of the original UK monetary contributions to be paid by the remaining memberstates. Additionally, some minor compensation payments received by other member countries (e.g.Sweden and the Netherlands) are also included in the calculations. Several alternative political power indices have been considered in the estimations, such asBanzhaf, Johnston, Public Good, and Deegan-Packel indices, see Kurz (2014) for a recent overviewon power indices. However, none of these power indices improved the explanatory power shown bythe Shapley-Shubik or the nucleolus. The Banzhaf and the Johnston indices show similar levels ofthe adjusted R while there are plenty of independent variables that are not significant. The PublicGood and the Deegan-Packel indices seem to be more sensitive to changes in the model specification. Budget shares have been calculated using the information taken from the European Commissionfinancial reports. The remaining data were taken from the Eurostat statistics website. Politicalpower indices have been calculated as described in Appendix 1.
Variable Mean Std. Dev. Min. Max. exp 0.0583 0.0530 0.0002 0.2256exp adj ssi nucl
One may observe that the SSI shows higher dispersion levels as compared to thenucleolus. The average expenditure budget percentage received is around 6%. It isalso worth mentioning that the variables representing budget needs present high levelsof dispersion. Thus, country members are heterogeneous in terms of their economicstructure.
In order to carry out a sensitivity analysis, we have proposed four different speci-fications. Estimates appear in Tables 3-7. The four specifications are the result ofcombining two different dependent variables ( exp in Tables 3 and 4; exp adj in Tables5 and 6) with the two political power indices described earlier ( p ssi in Tables 3 and 5; p nucl in Tables 4 and 6). Finally, Table 7 provides the basis for a comparison of thefour specifications, as marginal effects are reported.Regarding the econometric techniques, we have considered several models. First,we keep the pooled baseline Ordinary Least Squares specification (OLS) originallyproposed by Kauppi and Widgr´en (2004), in order for it to be compared with moresophisticated techniques. The analysis presented in the current paper suggests un-observable heterogeneity due to the strong differences among country members fromdifferent perspectives. Moreover, since the dependent variable is a share, economet-13ic methods should be adapted to take this into account. Thus, OLS seems to be anon-robust econometric technique in this context.In order to resolve any issues associated with the OLS method, two fractionalmethodologies have been proposed. First, a Generalized Linear Model (GLM) basedon a probit distribution has been applied. Second, since our panel data is clearly un-balanced, an alternative fractional model based on probit distribution has been con-sidered (FHETPROB). Note that the nature of unbalancedness could require modelsthat explicitly allow for heteroskedasticity (Wooldridge, 2010 a , b ). In both cases, aclustered option has been used to estimate the variance − covariance matrix.Additionally, Equation (1) has been extended to include a set of variables thatconsider the effect of the EU enlargement ( EU ). In this respect, a set of dummyvariables has been defined: EU EU EU EU EU
27. Those variablestake the value of 1 when the number of country members is 10, 12, 14, 25 and27 respectively, and 0 otherwise. Furthermore, some interactions of these dummyvariables with the political power indices have been considered, since there could bedifferent impacts of power depending on the number of countries integrated into theEU. The dummy variables have been included in OLS (denoted by
OLS d ), GLM andFHETPROB models.Moreover, additional dummy variables have been generated (Wooldridge, 2010 b )in order to capture the panel unbalancedness structure in the context of the FHET-PROB specification. Assuming that the global panel data set is composed of differentsubpanels T i , both the outcome and variance equation are allowed to depend on thenumber of observations in each subpanel. The new dummy variables are denoted by tobs tobs tob tobs tobs
6, and take the value 1 when the country is observedfor 32, 27, 18, 9 and 6 years respectively, and 0 otherwise.The results show some general facts that are observed in the majority of cases.Both power and needs are significant key drivers of budget allocation. Thus, thehigher the political power, the higher the expenditure share. Additionally, those14ountries with more intensive agricultural activity and lower relative income emergeas the beneficiaries of EU policies, as they receive higher shares of the overall budget.Regarding the econometric models, and comparing both fractional techniques, itseems that FHETPROB enables an encreased significance of the three main variablesrelated to power and needs. The inclusion of a variance equation based on the unbal-ancedness structure helps to refine the results. In all the specifications, the majorityof variables explaining the variance are highly significant.When comparing the performance of different models and when focusing on alter-native power indices, there are also interesting findings. In terms of OLS regressions,the adjusted R values show that models based on p ssi have a higher explanatorypower than those based on p nucl . However, the differences are considerably smallerwhen dummy variables representative of EU enlargement process are included in theanalysis. Thus, both power indices seem to perform similarly. Differences in termsof information criteria (aic, bic) show that fractional models using different powerindices are also very close to each other. Comparing Tables 5 and 6, the highest lagin information criteria between both power indices is registered for the specificationwhere adjusted budget shares are explained and heteroskedascitity is modelled. Inthis particular case, the nucleolus performs better.Regarding enlargement, two interesting effects are noted. First, the EU dummyvariables show that budget shares decrease with the EU size. Thus, the larger thenumber of country members, the lower the average budget share. Second, powerinteractions with temporal dummies show that there are different impacts of powerin different subperiods, especially when the nucleolus is included in the specifications.In general, the period 2004-2006 (EU25) is one where country members achieve higherrelative returns from their political power. When the number of country membersexperience a substantial increase, those countries with higher power levels obtainhigher relative gains.As mentioned earlier, Table 7 provides marginal effects, allowing the compari-15able 3: Total budget shares ( exp ) and the SSI ( p ssi ) (1) (2) (3) (4)OLS OLS d GLM FHETPROBp ssi 0.545 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.352 ∗∗ ∗∗ ∗∗ ∗∗ income -0.005 ∗ -0.004 + -0.157 + -0.263 ∗∗ p ssiEU10 0.034 0.490 0.659p ssiEU12 0.001 1.305 1.401p ssiEU15 0.199 ∗∗ ∗ ∗∗ ∗∗ ∗ ∗∗ ∗ -0.233 -0.103EU25 -0.014 ∗ -0.482 ∗ -0.142EU27 -0.010 + -0.450 ∗ -0.034tobs32 -0.569tobs27 0.540 ∗∗ tobs18 -4.400 ∗∗ tobs9 0.665 ∗∗ tobs6 0.918 ∗∗ cons 0.006 ∗ + -1.830 ∗∗ -1.722 ∗∗ lnsigma2tobs32 0.370tobs27 -0.419 ∗∗ tobs18 1.053 ∗∗ tobs9 -0.711 ∗∗ tobs6 -1.112 ∗∗ cluster no no yes yes N
575 575 575 575R adj 0.88 0.88chi2 580.92aic -2940.72 -2945.21 200.09 269.57bic -2923.30 -2884.25 261.05 356.66 + p < . ∗ p < . ∗∗ p < . exp ) and nucleolus ( p nucl ) (1) (2) (3) (4)OLS OLS d GLM FHETPROBp nucl 0.221 ∗∗ ∗∗ ∗ ∗ agri 0.504 ∗∗ ∗∗ ∗∗ ∗∗ income -0.010 ∗∗ -0.005 ∗ -0.168 + -0.277 ∗∗ p nuclEU10 0.037 0.471 + ∗ p nuclEU12 0.199 ∗∗ ∗∗ ∗∗ p nuclEU15 0.463 ∗∗ ∗∗ ∗∗ p nuclEU25 0.528 ∗∗ ∗∗ ∗ p nuclEU27 0.419 ∗∗ ∗∗ ∗ EU10 -0.009 + -0.108 ∗ -0.134 ∗∗ EU12 -0.026 ∗∗ -0.323 ∗∗ -0.374 ∗∗ EU15 -0.043 ∗∗ -0.507 ∗∗ -0.364 ∗ EU25 -0.043 ∗∗ -0.770 ∗∗ -0.388 + EU27 -0.039 ∗∗ -0.731 ∗∗ -0.280tobs32 -0.014tobs27 0.599 ∗∗ tobs18 -3.451 ∗∗ tobs9 0.705 ∗∗ tobs6 0.889 ∗∗ cons 0.022 ∗∗ ∗∗ -1.561 ∗∗ -1.454 ∗∗ lnsigma2tobs32 0.083tobs27 -0.483 ∗∗ tobs18 0.889 ∗∗ tobs9 -0.757 ∗∗ tobs6 -1.095 ∗∗ cluster no no yes yes N
575 575 575 575R adj 0.84 0.87chi2 641.14aic -2808.05 -2912.11 200.30 269.76bic -2790.63 -2851.15 261.26 356.85 + p < . ∗ p < . ∗∗ p < . exp adj ) and the SSI ( p ssi ) (1) (2) (3) (4)OLS OLS d GLM FHETPROBp ssi 0.768 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.200 ∗∗ ∗∗ ∗ ∗ income -0.002 0.001 -0.103 -0.223 ∗ p ssiEU10 0.034 0.456 0.493p ssiEU12 0.080 1.876 ∗ ∗ p ssiEU15 0.326 ∗∗ ∗∗ ∗∗ ∗∗ + p ssiEU27 0.210 ∗∗ ∗∗ ∗∗ -0.281 -0.153EU25 -0.012 + -0.490 ∗ -0.188EU27 -0.008 -0.443 ∗ -0.060tobs32 -4.990tobs27 0.490 ∗∗ tobs18 -5.168 ∗∗ tobs9 0.414 ∗∗ tobs6 0.931 ∗∗ cons -0.002 0.000 -1.926 ∗∗ -1.808 ∗∗ lnsigma2tobs32 1.454tobs27 -0.376 ∗∗ tobs18 1.163 ∗∗ tobs9 -0.491 ∗∗ tobs6 -1.091 ∗∗ cluster no no yes yes N
575 575 575 575R adj 0.87 0.88chi2 1121.58aic -2926.94 -2964.45 199.72 271.41bic -2909.53 -2903.49 260.68 362.85 + p < . ∗ p < . ∗∗ p < . exp adj ) and nucleolus ( p nucl ) (1) (2) (3) (4)OLS OLS d GLM FHETPROBp nucl 0.324 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.404 ∗∗ ∗∗ ∗ ∗ income -0.009 ∗∗ -0.001 -0.114 -0.251 ∗∗ p nuclEU10 0.039 0.477 + ∗ p nuclEU12 0.362 ∗∗ ∗∗ ∗∗ p nuclEU15 0.719 ∗∗ ∗∗ ∗∗ p nuclEU25 0.726 ∗∗ ∗∗ ∗∗ p nuclEU27 0.604 ∗∗ ∗∗ ∗∗ EU10 -0.010 + -0.114 ∗ -0.141 ∗∗ EU12 -0.041 ∗∗ -0.439 ∗∗ -0.493 ∗∗ EU15 -0.062 ∗∗ -0.646 ∗∗ -0.499 ∗ EU25 -0.055 ∗∗ -0.866 ∗∗ -0.505 ∗ EU27 -0.050 ∗∗ -0.810 ∗∗ -0.379 + tobs32 -0.217tobs27 0.589 ∗∗ tobs18 -3.597 ∗∗ tobs9 0.464 ∗∗ tobs6 0.890 ∗∗ cons 0.020 ∗∗ ∗∗ -1.574 ∗∗ -1.439 ∗∗ lnsigma2tobs32 0.193tobs27 -0.487 ∗∗ tobs18 0.909 ∗∗ tobs9 -0.547 ∗∗ tobs6 -1.083 ∗∗ cluster no no yes yes N
575 575 575 575R adj 0.81 0.87chi2 1283.98aic -2704.59 -2906.01 200.04 269.71bic -2687.18 -2845.05 261.00 356.80 + p < . ∗ p < . ∗∗ p < . In this respect, the original estimates of the SSI by Kauppi and Widgr´en (2004) for the period1976-2001 (considering annual data) ranged between 0.545 and 0.645 and between 0.783 and 0.858for the total budget shares and adjusted budget shares respectively. The impact of agriculture wasestimated in a range between 0.387 and 0.405 and between 0.252 and 0.236 for the total budgetshares and adjusted budget shares respectively. Finally, estimated income coefficient registeredvalues between -0.025 and 0.008 and between -0.022 and 0.003 for the total budget shares andadjusted budget shares respectively.
OLS OLS d GLM FHETPROBTotal budget share/SSIp ssi 0.545 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.352 ∗∗ ∗∗ ∗∗ ∗∗ income -0.005 ∗ -0.004 + -0.014 + -0.031 ∗∗ Total budget share/nucleolusp nucl 0.221 ∗∗ ∗∗ ∗ ∗ agri 0.504 ∗∗ ∗∗ ∗∗ ∗∗ income -0.010 ∗∗ -0.005 ∗ -0.014 + -0.033 ∗∗ Adjusted budget share/SSIp ssi 0.768 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.200 ∗∗ ∗∗ ∗ ∗ income -0.002 0.001 -0.009 -0.024 ∗ Adjusted budget share/nucleolusp nucl 0.324 ∗∗ ∗∗ ∗∗ ∗∗ agri 0.404 ∗∗ ∗∗ ∗ ∗ income -0.009 ∗∗ -0.001 -0.010 -0.029 ∗∗ + p < . ∗ p < . ∗∗ p < . The main contribution of this paper is to highlight the role of political power onthe EU budget decisions. Various key drivers of budget shares allocated to eachEU member country have been identified. Both power and needs are significantfactors in explaining expenditure budget allocation among EU member states. Someprevious empirical analysis (Kauppi and Widgr´en, 2004, 2007) show strong prevalenceof political power motives. Their results indicate that a large percentage of budgetexpenditures can be attributed to selfish power politics, leaving a small contributionto the so-called benevolent EU need-based budget policies.We have carried out an empirical analysis to revisit the findings of Kauppi and Widgr´en(2004). To this end, we have updated their data set (originally from 1976 to 2001,21he range has been extended to 2012). Additionally we have compared alternativepolitical power measures and have applied more sophisticated and rigorous econo-metric techniques. We have argued that the nucleolus (Schmeidler, 1969) is a goodalternative to the SSI when explaining the budget share of EU member states, fromboth a theoretical and an empirical perspective.Our findings show that under simple econometric specifications, both power in-dices behave in a similar way, although the SSI is slightly superior in terms of ex-planatory power. However, when using more sophisticated and adequate econometrictechniques, the nucleolus seems to perform better than the SSI. In particular, thenucleolus performs better when considering adjusted budget shares (by compensa-tions, such as the UK rebate), and when we adjust for the unbalancedness of thepanel data. Moreover, the higher the number of countries competing for EU budget,the higher the impact of political power on budget shares.Additionally, we find that the relative weight of political power based on the nu-cleolus when explaining budget shares is lower than predictions of other models. Apartial explanation may be the fact that the nucleolus can assign zero power to nondummy players, which is not the case for the Shapley-Shubik and the Banzhaf indices.Indeed, this occurs for all voting systems until 1994 (see Appendix 1). Needs alsomatter, and countries with lower relative income levels and a broader irrigation sectorare recipients of a significant amount of EU resources. These findings are consistentwith the idea that the EU budget is allocated to develop key policies such as thecommon agricultural policy and the structural funds. Although political power hasan impact on the EU budgetary decisions, this impact seems to be more moderatethan estimated in previous literature. Definitively, the solidarity principle emerges as In general, the voting power need not be proportional to the voting weights(Felsenthal and Machover, 1998). However, following a na¨ıve approach, we have also performedestimations using the voting weights instead of the power indices. The results imply that the spec-ifications based on the power indices perform better. More details are available from the authorsupon request.
22 significant key driver of the EU budgetary allocation.Finally, we would like to pursue this line of research further through a more focusedanalysis of specific sections of the EU budget. When modelling bargaining schemes,interactions among different sections of the EU budget will be considered. Addition-ally, further empirical analysis will aim to detect factors that increase the probabilityof receiving higher budget shares for specific policies.
Acknowledgements
We are grateful to professor Heikki Kauppi, who shared the original data set (1976-2001) with us. We would also like to thank Michel Le Breton, Mar´ıa Montero, Fran¸coisSalani´e and Roberto Mart´ınez Espi˜neira for their input during insightful discussionsat the early stages of our work, and the Editor and two anonymous referees for theircomments and suggestions.
References
Aksoy D (2010). Who gets what, when, and how revisited: Voting and proposalpowers in the allocation of the EU budget,
European Union Politics (2), 171–194.Anderson K and Tyers R (1995). Implications of EC expansion for European agri-cultural policies, trade and welfare, in R Baldwin, ed.,
Expanding the Membershipof the European Union , Cambridge University Press.Baron D and Ferejohn J (1989). Bargaining in legislatures,
American Political ScienceReview , 1181–1206.Bindseil U and Hantke C (1997). The power distribution in decision making amongEU member states, European Journal of Political Economy (1), 171 – 185.23luhm R (2013). fhetprob: A fast QMLE Stata routine for fractional probit modelswith multiplicative heteroskedasticity, Technical report.Courchene T et al. (1993). Stable money-sounds finances: Community public financein the perspective of EMU, European Economy .Felsenthal D and Machover M (1997). The Weighted Voting Rule in the EU’s Councilof Ministers, 1958-95: Intentions and Outcome, Electoral Studies (1), 33–47.Felsenthal D and Machover M (1998). The Measurement of Voting Power. Theoryand Practice, Problems and Paradoxes , Edward Elgar, Cheltenham.Gul F (1989). Bargaining Foundations of Shapley Value,
American Journal of PoliticalScience (57), 81–95.Hart S and Mas-Colell A (1996). Bargaining and Value, Econometrica (2), 357–380.Heinemann F (2003). The political economy of EU enlargement and the Treaty ofNice, European Journal of Political Economy (1), 17 – 31.Kandogan Y (2000). Political economy of eastern enlargement of the European Union:Budgetary costs and reforms in voting rules, European Journal of Political Economy (4), 685 – 705.Kauppi H and Widgr´en M (2004). What determines EU decision making? Needs,power or both?, Economic Policy (39), 221–266.Kauppi H and Widgr´en M (2007). Voting rules and budget allocation in the enlargedEU, European Journal of Political Economy (3), 693–706.Kurz S (2014). The inverse problem for power distributions in committees, p. 46.arXiv preprint http://arxiv.org/abs/1402.0988, submitted.24e Breton M, Montero M and Zaporozhets V (2012). Voting Power in the EU Councilof Ministers and Fair Decision Making in Distributive Politics, Mathematical SocialSciences , 159–173.Maschler M, Peleg B and Shapley L (1979). Geometric properties of the kernel, nucle-olus, and related solution concepts, Mathematics of Operations Research (4), 303–338.Maschler M, Solan E and Zamir S (2013). Game Theory , Cambridge University Press.Montero M (2005). On the Nucleolus as a Power Index,
Homo Oeconomicus
XXII (4), 551–567.Montero M (2006). Noncooperative bargaining foundations of the nucleolus in ma-jority games,
Games and Economic Behavior , 380–397.Montero M (2013). On the Nucleolus as a Power Index in Power, Voting and VotingPower: 30 Years After, edited by M. Holler and H. Nurmi, Springer, pp. 283–299.Napel S and Widgr´en M (2004). Power measurement as sensitivity analysis, Journalof Theoretical Politics , 517–538.Schmeidler D (1969). The nucleolus of a characteristic function game, SIAM Journalon Applied Mathematics , 1163–1170.Shapley L and Shubik M (1954). A method for evaluating the distribution of powerin a committee system, American Political Science Review , 787–792.Tangermann S (1997). Reforming the CAP: A prerequisite for eastern enlargement, in H Siebert, ed.,
Quo Vadis Europe? , Springer, pp. 151–179.Vidal-Puga J J (2008). Forming coalitions and the Shapley NTU value,
EuropeanJournal of Operational Research (3), 659–671.25ooldridge J (2010 a ). Correlated Random Effects Models with Unbalanced Panels,Technical report, Michigan State University.Wooldridge J (2010 b ). Econometric Analysis of Cross Section and Panel Data , MITPress. 26 ppendix 1: The SSI and the nucleolus in the Council of Ministers 1958-2002
Country 1958 − − − − − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . − − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . − − . . . . . . − − . . . . . . − − − − . . . . . − − − − − − . . . . − − − − − − . . . . − − − − − − − − . . − − − − − − − − . . − − − − − − − − . . S ource: Adapted from Le Breton et al. (2012). ppendix 1 (cont.): The SSI and the nucleolus in the Council of Ministers under the Treaty of Nice,2003-2012 Country 2003 2004 − − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − . . . . − − − − . . − − − − . . S ource: Own elaboration. ppendix 2 In this section, we introduce basic notions commonly used to model voting situations,and then briefly discuss the nucleolus and the Shapley-Shubik index.We consider a set N = { , ..., n } of n players or voters, which is often referred toas an assembly. The power set 2 N contains all the subsets of N . A non-empty subset S ⊆ N is called a coalition. The coalition N is said to be the grand coalition.A cooperative game with transferable utility in characteristic function form, is apair ( N, v ) with N the set of players and: v : 2 N −→ R : S v ( S ) , which is a map that satisfies v ( ∅ ) = 0. The map v is called the characteristic function.The value v ( S ) is said to be the value or the worth of coalition S . For simplicity, werefer to these games as “games in TU form.”The game ( N, v ) is said to be simple if: · the value of a coalition is either 0 or 1: v ( S ) ∈ { , } for all S ⊆ N , · the value of grand coalition is 1: v ( N ) = 1.A coalition with a value equal to 1 is said to be winning, and a coalition witha value equal to 0 is said to be losing. A winning coalition S is minimal if it doesnot contain any other winning coalition: v ( S ) = 1 and v ( T ) = 0 for all T ⊂ S .Furthermore, the set of winning coalitions is denoted by W and the set of minimalwinning coalitions is denoted by W m . The simple game ( N, v ) is fully determinedthrough the pair ( N, W ).Furthermore, the simple game is said to be monotonic if supersets of winningcoalitions are winning, i.e., if S ∈ W and T ⊃ S , then T ∈ W . A monotonic simplegame is also called a simple voting game. Several authors also use the term simple game for simple voting games, i.e., where monotonicity
N, v ) is said to be a weightedmajority game if there is an n -tuple w = ( ω , ..., ω n ) of non-negative weights with ω + ω + ... + ω n = 1 and a non-negative quota q such that v ( S ) = 1, if and only if,the total weight of the players in S exceeds the quota q , i.e., v ( S ) = 1, if and only if, X i ∈ S ω i ≥ q. The pair [ q ; ω ] is called a representation of the game ( N, v ). Typical examples ofweighted majority games are: · the majority game: w = (1 , , ..., | {z } n ) and q = ( n + 1) / , · the unanimity game: w = (1 , , ...,
1) and q = n , · the dictator game: w = (1 , , , ...,
0) and q = 1 (player 1 is the dictator).A measure of power is a map ξ from the set of simple voting games ( N, v ) to theset of n -tuples of real numbers. The value ξ i = ξ i ( N, v ) is the power of player i inthe game ( N, v ) , and it satisfies 0 ≤ ξ i ≤ One of the most famous power measures used in the literature is the Shapley-Shubikindex. Several approaches are used in the literature to present and interpret theShapley-Shubik index. Shapley and Shubik (1954) apply the following scheme tointroduce their index. The players vote in order and as a majority is reached, the billis passed. The critical voter is assumed to be given credit for having passed the bill. is assumed. For the definitions and properties, see Felsenthal and Machover (1998). Player i in coalition S is said to be critical in S if without player i the coalition left behind isloosing, i.e., i is critical in S if i ∈ S ∈ W and S \ { i } / ∈ W . If i is not critical in any S ∈ W , then i is a dummy. Definition 1
Let ( N, v ) be a simple voting game. The Shapley-Shubik index (SSI)of player i is defined by: φ i = φ i ( N, v ) = X S : i is critical in S ( | S | − n − | S | )! n ! for all i ∈ N . (2)The advantage of this approach is that it is simple and non-technical. However,the authors emphasize that this scheme, “is just a convenient conceptual device.”The main shortcoming of this scheme is that this voting model cannot be consideredrealistic: there is no reason why the pivot voter should get all the credit, or whythe order of the grand coalition formation should matter. With respect to thecomputation the Shapley-Shubik indices of our EU instances, we remark that loopingover all 2 n coalitions and determining the critical voters was quick enough for ourpurpose, i.e., more advanced methods involving generating functions were not needed. The nucleolus is a solution concept for cooperative games, which was first formulatedby Schmeidler (1969). In order to introduce it let us consider a characteristic functiongame (
N, v ). For convenience, for some vector x we define: x ( S ) ≡ X i ∈ S x i for any S ⊆ N. A payoff vector x = ( x , ..., x n ) with x i ≥ v ( i ) and x ( N ) = v ( N ) is called animputation. We denote by X ( N, v ) the set of all imputations of the game (
N, v ).Let x be an imputation, then for any coalition S the excess of S is defined as: e ( S, x ) = v ( S ) − x ( S ) . One might interpret this number as a measure of “dissatisfaction” for coalition S at imputation x . For any imputation x let S , ..., S n − be an ordering of the For more detailed discussion, see Felsenthal and Machover (1998). e ( S l , x ) ≥ e ( S l +1 , x ) for l = 1 , ..., n −
2. Let E ( x ) be the vectorof excess defined as E l ( x ) = e ( S l , x ) for all l = 1 , ..., n −
1. We say that E ( x ) islexicographically less than E ( y ) if: E l ( x ) < E l ( y ) for the smallest l for which E l ( x ) = E l ( y ) . We denote this relation by E ( x ) ≺ lex min E ( y ). Definition 2
The nucleolus is the set of imputations x for which the vector E ( x ) islexicographically minimal: ν = ν ( N, v ) = { x ∈ X ( N, v ) : ∄ y ∈ X ( N, v ) : E ( y ) ≺ lex min E ( x ) } . The following recursive procedure is used to characterize the nucleolus. By defini-tion, E ( x ) is the largest excess of any coalition relative to x . At the first step of theprocedure we find the set X of all imputations x that minimizes E ( x ):min ǫ s.t. e ( S, x ) ≤ ǫ for all S, ∅ " S " N and x ( N ) = v ( N ) . . The set X is called the least core of c . If it is not a unique point, we findthe set X of all x in X that minimizes E ( x ), the second largest excess and soon. This process eventually leads to an X k consisting of a single imputation, calledthe nucleolus (Schmeidler, 1969; Maschler et al. , 1979). The nucleolus recursivelyminimizes the ”dissatisfaction” of the worst treated coalitions. It has been proved that the nucleolus of a game in coalitional form exists and it isunique. Moreover, if the core is not empty, the nucleolus is in the core (Maschler et al. ,2013). Notwithstanding the general recursive definition of the nucleolus, the computation of X wassufficient in all EU instances, i.e., a single linear program has to be solved. The uniqueness of thesolution was verified using the complementary slackness condition.wassufficient in all EU instances, i.e., a single linear program has to be solved. The uniqueness of thesolution was verified using the complementary slackness condition.