The European Parliament (EP) is viewed as a normal parliament. Voting patterns of its members (MEPs) are mainly aligned with transnational political groups, not national cleavages. However, to label any parliament as normal requires its members to be responsive to the constituencies. But how can the responsiveness of the MEPs be measured and whom do they need to respond to? The intuitive answer to the latter question is that the MEPs view their countries as their constituencies. Logically, the very fact that the MEPs usually vote with their European political groups (EPGs) and not along national lines points to the lack of responsiveness on their part.
Yet, national affiliations matter. Hix (2005) admits that “party cohesion can break down when national interests are at stake”. The metaphor of the MEPs being agents of two principals, that is of their national party and their transnational political group in the EP, is especially revealing in explaining such situations. With re-election assumed to be the primary goal of a MEP, the crucial role that national parties play in nomination procedure will simply outweigh the importance of the EPGs in distributing offices and legislative reports, thus shifting loyalty away from the EPGs to national parties and making the MEPs vote among the national lines.
As a result, voting patterns of the MEPs are an outcome of possibly conflicting pressures. This makes voting data a distorted indicator of MEP responsiveness to constituencies. Instead, in this study we rely on written questions that the MEPs address to the EU institutions to measure the responsiveness. When a MEP poses a question s/he is under less pressure from the two principals than in case of voting. Using the universe of more than 100,000 such questions in 2002-2015 linked with MEP national and EPGs affiliation data, we test whether the issues of high sensitivity to their domestic audiences make the MEPs take their nationality seriously and pay more attention to those issues no matter what their partisan affiliations are.
We rely on supervised machine learning to uncover sentiment of every question asked in the EP on a negative-positive scale. Then we contrast the sentiment of questions related to Russia with the rest of questions conditional on party and national affiliation of the MEP asking the question. Our preliminary results indicate that (i) MEP questions involving Russia are significantly more negative in tonality, (ii) more variation in negative modality of Russia-related questions is explained by MEP national affiliation than his/her EPG. Our findings are robust to alternative methods of sentiment extraction and account for time-invariant unobserved heterogeneity of MEPs.