Analysis of dichotomous outcomes: a case study of constituency characteristics’ data in the 2006 general elections in Uganda

Abstract

Using election data on the 2006 Presidential poll following re-introduction of multi-party system and various surveys’ data compiled around 2006, this study explored characteristics that could have guaranteed election victory to the incumbent. The study considered scores for Constituency ratings – derived or otherwise - on income and poverty levels, participation, incumbency, literacy levels, party dominance or allegiance of the Constituency to a party, effect of regional location and the nature of the Constituency. The study population comprised 214 Parliamentary Constituencies covering the whole of Uganda.

A systematic approach utilising Hosmer-Lemeshow logistic regression technique for analysis of dichotomous outcomes was used. On analyzing the variables, all other quantitative predictors, except participation of the Constituency voters, average monthly per Capita incomes of voters and dominance by a political party, were eliminated. These were found to be statistically reliable for the restricted model in correctly predicting 70.56% of the data.

The policy implications for multiparty systems suggest that political and coalition actors could optimise their decisions or vote chances by articulating alternative policy matters in their campaign strategies. Unobservable income and other forms of inducements or coercion on which data are seldom collected or lacking may influence polls’ outcomes. Future studies could focus on the influence of these factors on election results. Technically, it remains incumbent upon the Electoral Commission and government to be held accountable for the entire electoral environment during the poll’s cycle.