Does remuneration affect the discipline and the selection of politicians? Evidence from pay harmonization in the European Parliament

For standard models of rent-seeking, see Tullock (1967), Buchanan et al. (1980), and Stigler (1971) for regulation. See Congleton et al. (2008) for an overview of research on rent-seeking. For a review of political shirking, see Bender and Lott (1996).

For a review of general ideas on the importance of selection in politics, see Fearon (1999), Brennan and Hamlin (2000), Besley (2005) and Mansbridge (2009). Recent research presents evidence showing that the identity of politicians matters for political outcomes, e.g., Chattopadhyay and Duflo (2004) and Jones and Olken (2005). An early contribution on the importance of politicians’ occupational backgrounds is presented by Couch et al. (1992).

Survey-based exceptions in political science are Scarrow (1997), Norris and Franklin (1997), and Hobolt and Hoyland (2011). Meserve et al. (2012) explore determinants of the party list positioning of candidates in the 2009 European election for 12 countries.

The absence rate as a proxy for shirking is used by Lott (1987) for the United States, Gagliarducci et al. (2010) for Italy and Besley and Larcinese (2011) for the United Kingdom.

Early work on the selection of national senior bureaucrats and national politicians into the EC Commission, based on reputational or budgetary considerations, is carried out by Vaubel (1994). A public choice approach to the study of international organizations, in particular the EU, highlighting the severe principal-agent problems is presented by Vaubel (2013).

An early contribution which discusses salary as a disciplining device in politics is Barro (1973).

Further institutions that have been addressed recently include the regime type (Besley and Reynal-Querol 2011), electoral rules and parties (Galasso and Nannicini 2011; Mattozzi and Merlo 2011), institutions enhancing transparency (Gehlbach et al. 2010; Rosenson 2006), and institutions which govern dual office-holding in different branches of government (Braendle and Stutzer 2010, 2011). For a first review, see Braendle (2013).

Whether higher pay impacts the incentive to rerun for political office is addressed by Groseclose and Krehbiel (1994) and Keane and Merlo (2010). Post-office earning possibilities are addressed by Eggers and Hainmueller (2009) and Parker and Parker (2009).

Estonia, France, Germany, Greece, Hungary, Spain, Portugal, Romania and the United Kingdom have a closed-list proportional representation system. The other member states have an open-list system. Ireland and Malta have a STV system in place (see Hix 2004). Rules of incompatibility which prohibit simultaneous dual office holding in the EP and the (sub-)national political assemblies apply in 22 member states (Lehmann 2009). Austria, Czech Republic, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Poland, Romania, Slovakia and Sweden have minimum vote requirements for parties (Nohlen 2004).

See Mamadouh and Raunio (2003), Kaeding (2004) and Benedetto (2005) for the importance and the determinants of report allocation in the EP. Major determinants of rapporteurship allocation identified by the literature, such as country of origin, party affiliation and tenure are taken into account in our empirical analysis.

MEPs attending all plenary sitting days in 1 year earn approximately 18,000 euros in per-diem allowances.

The new statute allows incumbent MEPs to continue receiving the former, nationally based remuneration as long as they remain in parliament. However, we do not have information on those MEPs who opted to retain the old pay scheme. According to the parallel work by Fisman et al. (2013), only 33 MEPs opted for the old scheme. Fisman et al. (2013) identify only five out of these 33 members.

For the states that were EU members before the expansion in 2004, the harmonization implied a significantly larger increase than those that had taken place during the period prior to EU enlargement. Ideally, we would like to have information on salary development for earlier periods in order to better rule out a pre-treatment time trend in pay. However, such information is not available. The Netherlands and Ireland are the only member states that had experienced larger increases in the period before harmonization. In robustness checks, we controlled for the sensitivity of our results by excluding the Dutch and the Irish delegations, the delegations that entered later (Bulgaria and Romania), and the delegations that experienced reductions in pay (Austria and Italy). For all three exercises, we found similar results.

We used these further sources when information was available in English, French, German, Italian, Polish and Spanish. In some cases, we referred to Wikipedia which provided further sources or information.

Table A.1 in the Online Appendix provides more detailed information on the classification and Table A.2 in the Online Appendix shows corresponding summary statistics at the delegation level. The Online Appendix is available at http://​wwz.​unibas.​ch/​personen/​profil/​person/​bra/​

Given that the dependent variables, plenary speeches, written declarations and reports drafted are logarithmized, we estimate specifications of a semilogarithmic functional form. Accordingly, the coefficient for a continuous variable shows the percentage change in the untransformed dependent variable per one-unit change in the explanatory factor. However, this interpretation of the estimated coefficients does not hold for categorical variables. Therefore, we also report estimated mean coefficients (in square brackets) that are consistent, close to the unbiased results and follow the interpretation of coefficients for continuous variables (Kennedy 1981).

Examples of such factors might be time-invariant differences in national institutions or national delegation-specific political cultures that influence the engagement in legislative activities.

As we look at two periods, our fixed effects estimation approach in levels is analytically the same as looking at first differences.

OLS does not produce estimated values outside the range of 0–100.

Their empirical strategy, however, differs from ours regarding the precise observation period under study, the inclusion of controls, the level of clustering, as well as the use of fixed effects.

Another possible explanation is that if services provided to local constituencies are more important for reelection than parliamentary attendance, an increased value of office-holding after the reform might lead to the politician providing more services to local constituencies and thus to more parliamentary absenteeism.

As the reform was enacted in summer 2005, the results for the effect of remuneration on in-office effort might constitute lower limit estimates. MEPs might have started to alter their behavior directly after the 2005 enactment. Such anticipation effects would lead to an attenuation bias.

The results of the analysis of sample restrictions, however, have to be interpreted with caution. We identify the effect of pay on the basis of a smaller number of observations and, in particular, rely on the variation in pay for (far) fewer than 27 countries, which makes the computation of (robust) standard errors less reliable.

We are aware of the challenges that come with the linear probability model as regards heteroscedasticity, distribution of errors and linearity. Therefore, we also ran the regressions with a logistic regression model. Applying a logit model causes a loss of observations owing to sparseness and the interpretation of the estimated coefficients is less straight forward. As the results generate coefficients of similar size when computing either marginal effects at the mean or average marginal effects, we prefer the linear probability model.

Regarding the parallel contribution by Fisman et al. (2013), we apply the same empirical strategy to, basically, the same data. Differences in the results for the part of the analysis that overlaps with theirs might, however, be because of the definition of the dependent variables, the inclusion or exclusion of countries (for the estimations on selection, we include Bulgaria and Romania), control variables (for instance, we include political group size, national delegation size, party affiliation and whether or not an MEP is chairing a committee or a delegation) or the precise legislative years under study.

In supplementary analyses, we investigated whether our results on political selection are driven by sample selection. First, in contrast to the finding of Kotakorpi and Poutvaara (2011), no evidence is found for systematic differences between the selection effects among women and men when looking at formal education (see Table A.4 in the Online Appendix, columns I and II). Second, as the potentially higher attractiveness of national parliamentary seats, in terms of reputation and decision-making power, might dissuade high-quality candidates from running for an EP seat, our results might be driven by the countries in which national elections take place in the same year as the EP elections. Excluding these countries, we find no different pattern (see Table A.5 in the Online Appendix). In the older member states with established selection procedures into the EP, the pool of candidates might react more strongly to a pay reform, as a seat in the EP is an established and relevant career option. For the corresponding sample restriction, we find more pronounced positive (education, Ph.D. degree) and negative (previous political experience) selection effects (see Table A.6 in the Online Appendix).

An early contribution on lawyers as legislators is presented by McCormick and Tollison (1978). They argue that lower parliamentary pay (indicative of a part-time parliament) attracts more lawyers as it is more compatible with their outside activities. However, the focus of their analysis is on the determinants of parliamentary pay.