An Easy and Accurate Regression Model for Multiparty Electoral Data

Katz and King have previously proposed a statistical model formultiparty election data. They argue that ordinary least-squares(OLS) regression is inappropriate when the dependent variablemeasures the share of the vote going to each party, and theyrecommend a superior technique. Regrettably, the Katz–Kingmodel requires a high level of statistical expertise and iscomputationally demanding for more than three political parties.We offer a sophisticated yet convenient alternative that involvesseemingly unrelated regression (SUR). SUR is nearly as easyto use as OLS yet performs as well as the Katz–King modelin predicting the distribution of votes and the compositionof parliament. Moreover, it scales easily to an arbitrarilylarge number of parties. The model has been incorporated intoClarify, a statistical suite that is available free on the Internet.