Fig. 5: Participants report that the thumbs up/down scale would lead to less expression of personal opinions, a greater focus on good versus bad quality, and would capture biased opinions less than using multipoint scales.
From: Scale dichotomization reduces customer racial discrimination and income inequality
a–c, Bar graphs showing mean perceptions of expression of personal opinions (a), a focus on good versus bad (b) and capturing biased opinions (c) by scale condition in study 3; n = 1,435. Error bars represent 95% CI. The mean expression of personal opinions of the thumbs up/down scale was significantly lower than the five-star scale (F(1, 1,431) = 6.65, P = 0.010, ηp2 = 0.005), the six-point scale (F(1, 1,431) = 25.62, P < 0.001, ηp2 = 0.02) and the ten-point scale (F(1, 1,431) = 41.37, P < 0.001, ηp2 = 0.03). The mean focus on good versus bad of the thumbs up/down scale was significantly higher than the five-star scale (F(1, 1,431) = 39.34, P < 0.001, ηp2 = 0.03), the six-point scale (F(1, 1,431) = 79.55, P < 0.001, ηp2 = 0.05) and the ten-point scale (F(1, 1,431) = 90.09, P < 0.001, ηp2 = 0.06). The mean capturing of biased opinions of the thumbs up/down scale was significantly lower than the five-star scale (F(1, 1,431) = 9.12, P = 0.003, ηp2 = 0.006), the six-point scale (F(1, 1,431) = 7.54, P = 0.006, ηp2 = 0.005) and the ten-point scale (F(1, 1,431) = 5.18, P = 0.023, ηp2 = 0.004). All tests were two-sided.
