In the 1950s, Chikio Hayashi developed an election prediction model for Japan by using sampling theory and face-to-face surveys. This model is built on the following steps: (1) defining the support rate for each candidate; (2) estimating the ratio of votes garnered by each candidate; (3) using Hayashi’s quantification method type 1 and cor- recting the difference between the value estimated from the third order regression and the ratio of votes; and (4) calculating the winning probability for each candidate by us- ing quantification method type 2 or the margin of error for the estimated ratio of votes. In the 1990s, the telephone replaced the face-to-face survey mode, providing changes to Hayashi’s election prediction model. We discuss the reasons for these improvements.
