THE FUNDAMENTAL BUILDING BLOCKS of political parties are candidates running for election in electoral districts. In the previous chapter we showed that the most common outcome in single-member, simpleplurality district elections in Canada, Great Britain, India, and the United States is for two candidates to garner a very large proportion of the vote in a district. Even in highly diverse districts, such as those in India, with many identifiable groups of voters differentiated by language, class, religion, and caste, it is typically the case that only two candidates compete seriously for election. These outcomes indicate that, more often than not, voters are coordinating on specific candidates, even if the mechanisms for that coordination are not easy to detect or describe. Somehow, candidates who are not expected to finish first or second in the balloting lose votes in droves or drop out, at least in most districts most of the time.
Theoretical models of Duverger's Law for district elections draw conclusions that are consistent with these findings. The models, summarized in chapter 2, conclude that in nearly every circumstance two candidates should receive all the votes. To reach this conclusion, the models rely to a certain extent on the notion of strategic, or fully rational, voting. In fact, coordination of voters on candidates essentially requires some form of strategic voting, because it means that some voters are implicitly agreeing to rally behind a candidate who may not be their first choice in the hopes of avoiding a bad outcome. According to Cox (1997, 72):
Strategic voting in a simple plurality election means voting for a lower-
ranked candidate that one believes is stronger, rather than for higher-ranked
candidate that one believes is weaker. … [If] voter beliefs about which
candidates are stronger and weaker will be generally correct, [then] strate-
gic voting will generally transfer votes from objectively weaker (vote-
poorer) to objectively stronger (vote-richer) candidates.
Many district elections among the cases discussed in chapter 2, however, do not conform to the expectations of Duverger's Law. The exceptions may require some reevaluation of models of Duverger's Law, but their existence does not have a bearing on our argument. Our interest lies in the formation of national party systems and whether the same