Does the Confidence in the EU Spill Over to the National Level? A Longitudinal Analysis of Political Trust in Central Europe

Abstract

The EU enlargement brought about a series of institutional changes in the new member-states. This article shows how the EU also shapes citizens' attitudes towards their domestic political institutions. Using survey data from the Candidate Countries and Standard Eurobarometers (2002–06), we show that political trust in national institutions in Central Europe is the effect of trust in the EU and of the expectations projected onto the evolution of the national economies. Such determinants got stronger after EU accession, thus supporting our central argument, according to which we witness a third phase in the formation of attitudes towards institutions in post-communist Europe. Keeping to these lines, our study sheds new light on the patterns and origins of political trust formation, showing that a transfer of legitimacy is possible not only from a national to a supranational level, but also in the reverse direction.

Key Words:

• Trust
• national political institutions
• EU
• Central Europe

Acknowledgments

Parts of this work were carried out at the European Data Laboratory for Comparative Social Research (EUROLAB). Access to the EUROLAB was supported by the European Community under the ‘Structuring the European Research Area' specific programme Research Infrastructures Action in the 6th Framework Programme.

Notes

   ¹ Extensive literature identifies strong connections between political trust and the functioning of democracy. Two mechanisms are at work. First, the scholarly tradition emerging from Almond and Verba (1960) and Easton (1965) claims that (stable) democracies require high levels of political trust as a form of diffuse support from their citizens; political trust legitimizes the government and enhances its effectiveness (Gamson 1968, Scharpf 1999, Rudolph and Evans 2005, Marien and Hooghe 2010). Accordingly, loss of trust can reflect attitudes towards political institutions and procedures (Putnam 2002, Dalton 2004). Second, there are recent studies indicating that low political trust is beneficial for democracies. Giving voice to discontent and critically assessing the performance of institutions can lead to more responsive actions from governments (Nye 1998, Norris 1999, Hardin 2004, Inglehart 2008).

   ² The Standard Eurobarometers 62.0, 63.4 and 65.2.

   ³ Interpersonal trust can be either thick (based on strong ties of family or friendship) or thin (based on acquaintances or professional environment) (Putnam 2000). This distinction, although relevant for the discussion of trust, does not occupy a relevant position in our analysis regarding the attitudes towards political institutions.

   ⁴ We also consider that corruption and government performance are not the best measures for confidence in institutions. Although theoretically relevant as a determinant of institutional trust (Diamond 1999), corruption is captured by satisfaction with democracy (Bhalla 1997) and reflected in the prospective economic evaluations. Moreover, Newton and Norris (1999, 5) report the ambiguous evidence produced by corruption perceptions. Government performance is, on the one hand, also highly correlated with satisfaction with democracy and, on the other hand, loses relevance in our longitudinal study in which we check the robustness of attitudes before and after accession.

   ⁵ For many post-communist countries, 1999 marked the beginning of accession negotiations.

   ⁶ In some cases, the relationship can have a different direction. For example, in the context of poor knowledge of citizens about international issues (Huckfeldt et al.2005, Gherghina 2010), individuals may use proxies when they are called upon to make decisions on issues about which they have little or no information. Anderson (1998) finds that in a few Western European countries, citizens tend to look to their domestic political institutions (governments or political parties) when deciding whether to support European integration. Another example is the empirical finding that incumbent popularity and government performance foster public support for European integration (Franklin et al.1994). However, post-communist Europe presents different scenarios. There is low trust in national institutions and high political trust in the EU (Tanasoiu and Colonescu 2008, Gherghina 2010). Moreover, the high level of electoral volatility in most of these countries and the lack of continuity in government (Gherghina 2011) indicate general discontent with the incumbents, while political trust in the EU was still high. Summing up, the mechanisms identified for Western Europe do not function in Central Europe.

   ⁷ Trust in the EU goes beyond trust in the EU institutions. We created an index for the latter and, at the pooled level, the two variables are only moderately correlated, at an R of .548

   ⁸ It was not possible to test longitudinally also for the impact of the retrospective economic evaluations.

   ⁹ We control for the incumbent authorities through our case and time period selection: 2005 and 2006 were electoral years in the selected countries and thus citizens have different incumbents as reference points.

¹⁰ The content of discussions is unclear – we do not know whether the respondents discussed institutions positively or negatively. However, we aim to capture socialization through communication, rather than the social network of the individual and the impact it has on perceptions.

¹¹ For the operationalization of all the variables, see the codebook in the Appendix.

¹² We used maximum-likelihood factor analysis: Two factors that had an eigenvalue larger than 1 were retained. The first captures the trust in political institutions, while the second measures the trust in the ‘law and order’ state apparatus.

¹³ The complementary log-log link function from STATA and R (proportional odds logit) corresponds to the negative log-log function in the SPSS ordinal regression (PLUM). This link function is particularly appropriate when the distribution of cases on the dependent variable is left censored.

¹⁴ On the contrary, the Wald Test for Parallel Lines Assumption (obtained through the ‘autofit gamma’ command of gologit2) indicated that the final model does not violate the proportional odds assumption.

¹⁵ Prior to this, we checked for multicolinearity by running an OLS model and then the VIF command. The resulting levels of variance inflation are not problematic at all (mean VIF between 1.09 and 1.14, with a maximum of 1.31 for democratic satisfaction).

¹⁶ This option has the advantage of avoiding altogether the problem of auto-correlation or serial correlation.

¹⁷ The interpretation of results suffers from the fact that we could not present first differences based on simulations (or the graphical representations associated with them) since the package used (gologit2) is supported neither by ‘estsimp’ – the function part of Clarify, designed for STATA (King et al.2000) nor by R's Zelig (Imai et al.2008). Running first differences based on a simple ordered logit model would produce potentially biased results.