van de Wardt, M. (2017) “Explaining the effective number of parties: Beyond the standard model”, Electoral Studies 45, 44-54 .


Previous research dealing with the explanation of the effective number of parties (ENP) has suggested several explanatory factors, for instance electoral institutions or societal divides, that should account for the variation in the number of parties across democracies. In the course of this strand of research the “standard model” of party system density emerged. The model foregrounds the interaction between social heterogeneity, mostly measured by ethnic fractionalization, and the permissiveness of the electoral system, operationalized by median district magnitude. At this juncture, the model emanates from a positive effect of higher social heterogeneity on the ENP only if the permissiveness of the electoral system is sufficiently high. The paper by van de Wardt addresses potential shortcomings of the standard model by proposing two additional factors to be considered: voters’ ideological preferences on the left-right dimension and issue diversity on the party system agenda.


In the article, van de Wardt tests five hypotheses on the factors shaping the ENP empirically. He hypothesizes that social heterogeneity (H1), the issue dimensionality and issue diversity (H2 & H3) on the party system agenda only increase the ENP when district magnitude is high enough. Furthermore, H4 states that a more fractionalized voter distribution only leads to an increasing ENP when district magnitude is quite low, what is due to strategic voting. Finally, these four hypotheses should only be confirmed in highly institutionalized party systems (H5), i.e countries that were already democratic before the 1990s. The hypotheses are tested empirically on the basis of data including 696 national legislative elections from 79 countries, covering the years from 1945 to 2011. The author applies a pooled OLS model, calculating whether the marginal effect of the independent variables on the dependent variable ENP differs from zero for all observed values of district magnitude (the model makes use of the logged median district magnitude). With the exception of H2, the analysis generally finds support for all addressed hypotheses.


The article convincingly demonstrates, making use of a high number of cases, that there is good reason to take further factors for future models on the ENP into account. Especially the finding that all hypothesized effects were only observed in long-established democracies shows that the perspective of voters and their adaption to the supply of the party system agenda and the conditions of the electoral system definitely needs to be addressed in future research. However, the article in large parts retests the existing approaches concerning the explanation of the ENP without grappling with further difficulties of former research. This pertains, for instance, to the shortcomings of setting the differentiation between the effective number of electoral (based on their share of votes) and legislative parties (based on their share of seats in parliament) aside. Taking this differentiation into account would be important to shed light on voters’ reaction to the electoral system and the diversity of the party system agenda. So, while the effective number of legislative parties is directly affected by changes of the electoral system, the effective number of electoral parties is quite depending on the process of adaption of new electoral rules by voters and party elites.

Author: Julian Noseck in March 2017

Kommentare: 2
  • #2 (Dienstag, 04 Juli 2017 18:42)

    Thank you very much for your comment! I admit to not having pointed out that it is solely about the ENP based on votes.

    There are certainly difficulties concerning the use of national data in this respect. But how could a retest of previous work be carried out without leaning on the complexities of these models? This is surely also true with regard to the general statistical problems of macro-comparative analyses. However, in this particular case I do not know how the sample could be significantly expanded. Therefore, going for analyses at the district level is definitely the most adequate approach.

  • #1 (Montag, 03 Juli 2017 11:05)

    This note is unclear whether author uses ENPvotes or ENPseats.

    How credible is it to test any model like this using national data for ENPvotes, which is inherently averaged out on both ENP side and the explanatory side? Surely only local district level tests are credible now? And under 700 averaged out cases is not a lot to go on.