A Method for Election Forecasting

Abstract

The CHAMPSKNOW system was used to predict the 2016 Jamaican General Election. The questions asked were: what were the probabilities of the JLP or the PNP winning the election? Would the election be close? How many seats would each party receive? Integrated or mixed methods were used with quantitative and qualitative data. These data were macro-economic and demographic indicators (1962-2015) and regression was done to identify the significant predictors of party in power; general election results (1962-2011) were used to calculate the vote share and frequency of wins and seat type; type of CDF disbursements (2008-2015) and a regression was done to determine the significant predictors of MP victory; published polls (1962-2016) were used to identify the most accurate poll to guide the forecast; independent voter survey in marginal seats was used with vote share, frequency of wins and seat type to determine seat count; newspaper stories about the election were coded as positive and negative stories about the parties; the parties’ audio-visual ads were coded as positive, negative and contrasts ads. The final forecast gave the PNP a 55% chance of winning (with 35 seats) and the JLP a 45% chance (with 28 seats), a close election. The JLP won with 49.5 % of the popular vote (31 seats) to the PNP’s 49.1% (30 seats) because of the promised tax break. This wildcard was not assessed since voters tend to ignore campaign promises because they were usually unfulfilled.