Repercusiones

The analysis of this chapter is performed on roll-call votes. These are the votes that are officially recorded, in the form of minutes drafted by the Parliament’s offices. The records track how each MEP voted or abstained. Commonly, this voting procedure is called “public vote”, because a member of a parliament has no means of keeping the vote secret. In the American, British and in few other parliaments all the votes are roll-calls; however, in many other cases such as Italy, France, Germany, Belgium, Argentina, and Brazil some votes are public and recorded, while others are secret (Hug 2006: 24-33).

In the European Parliament the choice of voting procedure is defined by the EP Rules of Procedure, which states that only certain votes are required to be taken by roll-call¹⁴. However, a political group or at least 32 MEPs can request any vote to be taken by roll-call. Before the start of the current, 7^th European legislature it was also provided that it is compulsory to have RCVs for all the final legislative votes, as stated in the new Rule 166¹⁵. In addition to RCVs, two other voting procedures are possible: the “raise of hands” and “voice” vote where it is recorded only if the bill passes or not, and another type of electronic vote that records the aggregate number of yea/nay, but no information about the single vote. Roll-call votes represent slightly more than 30% of the whole universe of votes cast in the EP, thus potentially leading to

14 Namely, Rule 106(5), which relates to the vote on the election of the Commission; and Rule 107(5), which relates to a motion of censure in the Commission.

15 “When voting on any proposal for a legislative act, whether by way of a single and/or final vote, Parliament shall vote by roll call using the electronic voting system” (Hix et al.

2013: 15).

selection bias, as anticipated in the previous chapter. Therefore, at this point it is necessary to discuss how the academic debate over the reliability of such dataset has evolved in the years.

3.2.1. Are roll-call votes a reliable source of information?

From the early ‘90s scholars of MEPs’ voting behaviour started to focus almost exclusively on roll-call votes (Attinà 1990; Hix and Lord 1997;

Raunio 1996; Hix 1999), as already discussed in chapter 2 and in the previous paragraphs. In response to these studies, Carrubba and Gabel (1999) were the first to draw attention on the potential selection bias in using RCVs to study MEPs’ voting behaviour. First of all, it has to be clarified that it is impossible to empirically analyse the votes falling under the other voting procedures, because there are no records of those vote. In other words, voting behaviour in non-RCVs is, simply, not directly observable and analysable with the quantitative approach usually employed for RCVs. However, RCVs are a small subset of the whole universe of votes cast (roughly ⅓). Thus, only if RCVs represent a random sample of the whole universe of votes cast, then the causal inferences based on RCVs can be generalized as a reflection of MEPs’

voting behaviour. If a dependent variable of a research (e.g. intra-party cohesion, inter-party competition) is related¹⁶ to the rationale behind the RCV request, endogeneity problems are expected to emerge. In the light of this it may be assumed that RCVs are requested by EPGs for strategic reasons.

However, the extreme usefulness (yet unpaired by an equal reliability and completeness of information provided) of roll-call datasets is indisputable. Their popularity has increased over time, and “in recent years, statistical analysis of roll call votes (RCVs) has become the bread and butter of empirical research on legislative behaviour in democracies” (Hix et al. 2013: 2). From the early ‘2000s almost every student of EP legislative behaviour has faced, somehow, RCVs analysis – or their potential shortcomings. Those who employ RCVs usually base their analysis on the following assumptions: RCVs and votes falling under a different voting procedure do not vary in terms of importance; both are cast on the same policy issues, so no issue is voted too often with a certain type of vote; and finally, all the EPGs are equally likely to request RCVs, so no group requests them

16 Either positively, negatively or both - but in a way such as the two effects do not cancel out each other.

disproportionally more often than the other groups (Hix and Noury 2002, Hix et al. 2007)

Meanwhile, an increasing number of critical scholars (Carrubba et al.

2003, 2004, 2008, 2009; Thiem 2006; Hug 2006, 2009) claim to have concrete evidence to confute the assumptions made by scholars that use them to analyse legislative behaviour. First of all, according to them the majority of non-RCVs are cast on legislative voting (the most important votes), while most of the RCVs are cast on resolution and consultation votes (less important). Second, RCVs are cast disproportionally on specific issues. Finally, some EPG requests them more often than other groups. Consequently, inferences based on RCVs only, may be misleading and it seems necessary to look somewhere else to fully understand the dynamics of voting behaviour within the EP. Later, these authors rephrased their criticisms in clearer terms, pointing out not only the problems of endogeneity in party cohesion, but also the possibly misleading way scholars pooled data together to create the dataset (Carrubba et al. 2003: 6-9). Consequently, these authors switched their research focus from voting behaviour to the reasons why RCVs are requested. Some scholars have tried to provide an explanation to the real reasons behind a RCV request, relying mostly on theory, while others have applied methods based on game theory to model this kind of situation. Carrubba et al. (2008: 570-572) consider the RCV requests a consequence of high party cohesion. On the other hand, Thiem (2006: 17) see these requests as signalling strategies either for a group to highlight their pre-existing unity or another group’s coordination failure. In other words, cohesion – or lack of it – increases the likelihood of a RCV request.

Nevertheless, this debate is far from being settled, as demonstrated by the latest working paper written on the topic. Again, Hix et al. (2013) try to confute these critical scholars’ argument by suggesting that, according to the original methodology proposed (a comparison between final votes in the sixth and seventh EP, exploiting the new provision according to which all final legislative votes have to be publicly recorded), selection biases in roll call votes due to strategic choices are at best negligible¹⁷.

17 Convincing arguments have been presented on both sides of this academic debate;

however, no matter how refined may be the model employed to prove that voting behaviour bears no difference in both types of vote, the central doubt remains: how can a non-random, non-representative sample lead to unbiased results?