Keywords: Political behavior, participation and culture

According to Gordon and Segura (1997), arguably the most influential piece in the study of political knowledge across countries, political systems with multiple parties provide more opportunities for voters to be informed given that parties have the incentive to differentiate themselves and provide more and better information to the electorate. Yet, this Downsian (1957) logic assumes that voters are well-equipped to absorb and process an increasing amount of political information, which contrasts with studies in psychology that suggest that increasing the number of choices may lead individuals to become discouraged by the efforts necessary to make an informed decision, particularly when the stakes are high (Luce, Bettman and Payne 1997, Schwartz 2004, Huang and Zhang 2013). Experiments on consumer behavior also provide evidence that more options can be demotivating as people are less likely to purchase products or to engage in activities when given a larger number of options (Iyengar and Lepper 2000, Iyengar 2010), and works from political behavior find that large amounts of options and information can lead to voter indecision, inaction, and disengagement from politics (Cunow 2014, Lau and Redlawsk 2006, Rahn, Aldrich and Borgida 1994, Sniderman and Levendusky 2007). So, what explains the divergence between these two literatures?

I argue that the results presented by Gordon and Segura (1997) are driven by inexact evaluation and interpretation. More specifically, I reevaluate the modeling and interpretation of the results presented by the authors, and present evidence that the estimated positive effect of increasing the number of parties is driven by two reasons. First, the small number of countries combined with the authors' definition of multiparty systems as more than 2.5 effective number of parties and analysis of nested data as independent and identically distributed observations leads to a misleading evaluation of the results. By taking into account the clustering in the estimation of standard errors
and testing different alternatives to modeling the relationship between number of parties and individual knowledge (e.g. different cut-off, curvilinear relationship), I demonstrate the original findings suffer from a lack of robustness. Second and more importantly, I find that the estimated positive effect is driven by the arbitrary imputation strategy for “Don’t know” responses that largely penalizes individuals who do not provide answer, which systematically affects the average levels of knowledge across countries. In addition, I use a separate dataset of 85 national surveys in 46 countries collected by the Comparative Study of Electoral Systems (CSES) project to re-test the relationship between the number of political parties in a system and individuals' levels of political knowledge. The results again highlight the role of the imputation strategy as an artifact, and similar to other more recent studies, suggest that the relationship between number of parties and knowledge of the citizenry does not fall in any particular direction.