What Would Happen if We Were Better Informed? Simulating Increased Knowledge in European Parliament (Ep) Elections

Bibliographic Data

ID7997967
AuthorsYosef Bhatti (0000-0002-8112-486X, corresponding author)
Year2010
Volume46
Issue4
Pages391-410
Publication date2010-11-01
Peer ReviewedYes
Open AccessNo
TypeARTICLE
VenueRepresentation (JOURNAL)
Journal identifiersISSN: 0034-4893 • E-ISSN: 1749-4001
PublisherTaylor & Francis (PUBLISHER • GB)
DOI10.1080/00344893.2010.518078
OpenAlexW1980450744
LanguageEN
Citations received13
References cited32

What would happen if the public in a multiparty system became better or worse informed? This article examines the consequences of political knowledge on turnout and vote choices in three recent Nordic European Parliament (EP) elections. Only a moderate increase in political knowledge would result in more than 3 percentage point higher turnout. The relative support for the parties would also be affected if knowledge increased significantly, though less so than for turnout. Social democratic parties would be the big losers and among the winners would in particular be conservative parties. Though the party level changes appear larger than in two‐party systems, the predicted changes in relative power between the political blocs are at best moderate. ACKNOWLEDGEMENTS The data utilised in this article were originally collected by the 2004 European Election Study research group. This study has been made possible by various grants. Neither the original collectors of the data nor their sponsors bear any responsibility for the analyses or interpretations published here. The data are available from the homepage of the European Election Study (www.europeanelectionstudies.net) and from the Central Archive for Empirical Social Research (ZA) at the University of Cologne, Germany. An earlier version of the paper was presented at the ECPR Joint Sessions, Lisbon, Portugal, 14–19 April 2009, Workshop 16: political information, public knowledge and perceptions of reality. I thank the participants and Gábor Tóka for their generous and critical comments. A special thanks to Kasper M. Hansen for numerous comments and ideas along the way. The usual disclaimer applies. Notes 1. See also Claassen and Highton (2006 Claassen, Ryan L. and Highton, Benjamin. 2006. Does policy debate reduce information effects in public opinion? Analyzing the evolution of public opinion on health care. Journal of Politics, 68(2): 410–20. [Crossref] , [Google Scholar]) who find that the knowledge gap even increases under intense public debate. 2. Only Sweden differed slightly, since an 11‐point scale was used in the EES. However, this should be inconsequential for the results. 3. Before any estimation is carried out, I imputed values for missing data. The knowledge variable contains no missing values, since a 'don't know' here conveys substantive information (i.e., the respondent does not know the answer). However, on the demographical and the predisposition variables needed to carry out the simulation, up to 7% of the data is missing due to refusal or inability to answer the questions. The problem is dual. First, the missing data reduce the efficiency of our analyses. Second, and most serious, low‐knowledge respondents are most likely to report an invalid answer for the demographical and predisposition questions (Althaus 2001 Althaus, Scott L. 2001. "Who's voted in when the people tune out? Information effects in Congressional elections". In Communication in U.S. Elections: New Agendas, Edited by: Hart, Roderick P. and Shaw, Daron. Lanham, MD: Rowman and Littlefield. [Google Scholar]). 4. Multinomial logit models are the practically most appropriate models for the present purpose even though they assume independence of irrelevant alternatives (IIA). As Dow and Endersby (2004 Dow, Jay K. and Endersby, James W. 2004. Multinomial probit and multinomial logit: a comparison of choice models for voting research. Electoral Studies, 23(1): 107–22. [Crossref] , [Google Scholar]) argue, for applications in studies of voter choice, IIA is neither particularly relevant nor restrictive (apart from studies of candidate entry or exit from elections which is not the focus of the present study). Even in the case of violations of IIA, the estimated logit probabilities will be consistent. In addition, Hausman tests were calculated and resulted in only one rejection of the null of independence (the non‐voter category in Denmark) from 25 tests even when using the generous 0.10‐alpha level. 5. Note the model—like most other simulations (e.g. Althaus 1998 Althaus, Scott L. 1998. Information effects in collective preferences. American Political Science Review, 92(3): 545–58. [Crossref], [Web of Science ®] , [Google Scholar], 2002 Althaus, Scott L. 2002. Collective Preferences in Democratic Politics. Opinion Surveys and the Will of the People, Cambridge: Cambridge University Press. [Google Scholar]; Bartels 1996 Bartels, Larry M. 1996. Uninformed voters: information effects in presidential elections. American Journal of Political Science, 40(1): 194–230. [Crossref], [Web of Science ®] , [Google Scholar]; Hansen 2009 Hansen, Kasper M. 2009. Changing patterns in the impact of information on party choice in a multiparty system. International Journal of Public Opinion Research, 21(4): 525–46. [Crossref] , [Google Scholar])—does not include second‐order interaction terms due to multicollinearity concerns, i.e., when knowledge‐squared and all its interactions with demography are included the model, it tends to drown in multicollinearity. The current specification assumes that the direct knowledge effect is monotonic. However, by including all interactions with socio‐demographic characteristics, it is still quite flexible, since it allows for different magnitudes and directions of the information effects for different types of people, for each of the choices/parties (see Tóka 2002 Tóka, Gábor. 2002. "Voter inequality, turnout and information effects in a cross‐national perspective". Helen Kellogg Institute, Notre Dame University. Working Paper no. 297 [Google Scholar] for a thorough discussion of the up‐ and downsides of the different specification possibilities). For instance, the coefficients in Table 1 indicate a negative conditional knowledge effect of being a highly educated constituent on social democratic vote compared to a positive conditional effect of living in a rural area. Nevertheless, I did try disregarding the multicollinearity concerns and estimate models that included knowledge‐squared and all interactions with knowledge squared for a 0.25 SD increase in knowledge and a 0.25 SD decrease to test the sensitivity of the results to this alternative specification. As for turnout, the predicted changes from the two specifications were almost perfectly congruent and correlated above 0.99. Regarding prediction of party‐level changes (without turnout changes), the predictions of changes (i.e., the difference between two absolute predictions) were somewhat more different, but still correlated 0.54 (the absolute predictions of course correlated much closer). At the party group level, the correlation was 0.78 (probably because random noise produced by the multicollinearity tends to cancel out at the higher level of aggregation). 6. It is important to note that the simulations are based on the coefficients from the partial slopes—not from simple bivariate correlations. That is, if for instance the analysis shows that left parties would lose votes if the constituents became better informed; this implies that left‐wing voters on average are less informed given all other variables included in the model and conditional on all 11 interactions

Democracy · Parliament · Political economy · Political science · Politics · Power (physics · Sociology · Turnout · Voter turnout · Voting · Electoral Systems and Political Participation · Law · Political Influence and Corporate Strategies · Public Administration · Social Media and Politics

Tools

Open DOISci-Hub