To accurately predict the winner of a certain election, polls need to consider the opinions of the appropriate target audience, in this case it would be individuals most inclined to actually vote. This audience can be identified in two ways, the first is by asking the individual participate whether he/she plans on voting during the upcoming elections. The second option is to conduct a thorough study on the characteristics of voters in previous elections. Pollsters should, thus, ensure that the target audience of their polls match a similar profile to the ones who have historically voted in previous elections. This could be done through developing predictive models on citizen’s likelihood to vote. These models can be established through using multivariate logistic regressions. The analysts working on a certain poll to predict the winner in a given presidential election should calculate the prevalence of support for a certain candidate among those who show a high probability of voting in the upcoming election.
Comments