This paper analyses the empirical close-election strategy, which uses electoral races in single-member constituencies decided by a small margin of votes to generate a "quasi random" variation in the winner's identity and to study the effect of having elected a candidate of a given party on a certain variable (e.g. fiscal policy).
This paper argues that, in the absence of variation in the parliamentary majority during the sample period, the close-election identification strategy does not isolate the effect of electing a member of a given party from that of the latter's membership of the parliamentary majority or the opposition. It shows that this identification is possible only in the presence of variations in the parliamentary majority, and underlines the relevance of these results by replicating some estimates in the literature.
