Political knowledge reduces hindsight memory distortion in election judgements

Abstract

Hindsight bias occurs when outcome information biases judgements. Previous studies have demonstrated hindsight bias in judgements of election outcomes, but few studies have examined the role of domain knowledge in hindsight bias. The present study examined the relationship between political knowledge and hindsight bias using both memory and hypothetical designs. Participants answered political knowledge questions and some made predictions before the 2012 US Presidential Election. After the election, participants were provided with the outcomes. Those who made predictions were asked to recall them, whereas those who did not make predictions were asked what they would have predicted. Both groups demonstrated hindsight bias: their recalled or hypothetical predictions were closer to the election results than participants' actual predictions. Political knowledge was negatively correlated with hindsight bias in recalled predictions but not significantly correlated with hindsight bias in hypothetical predictions. These findings help elucidate the role of domain knowledge in hindsight bias.

Keywords:

• Domain knowledge
• Elections
• Hindsight bias
• Prediction

We thank Rafael Ayala, Michelle Calderwood, Araceli Cruz, Amanda Gochez, Eleuterio Limas, Emily Limas, Jesus Quintero and Joseph Schmitt for data collection, and Lawrence Becker for feedback on the political knowledge questionnaires.

We thank Rafael Ayala, Michelle Calderwood, Araceli Cruz, Amanda Gochez, Eleuterio Limas, Emily Limas, Jesus Quintero and Joseph Schmitt for data collection, and Lawrence Becker for feedback on the political knowledge questionnaires.

Notes

¹ When analysed separately, the Pew political knowledge questionnaire was significantly correlated with hindsight bias in the memory condition, r(69) = −.28, r = .019, whereas the McGlone et al. questionnaire was not: r(69) = −.08, p = .514.

² The hindsight index for each candidate did not significantly differ based on which candidate participants wanted to win the election in the memory condition: the hindsight index for Obama among those who wanted Obama to win (n = 44; M = −0.08, SD = 7.24) was not significantly different from those who wanted Romney to win (n = 15; M = 0.91, SD = 5.25), t(57) = 0.49, p = .629, d = 0.16; the hindsight index for Romney among those who wanted Obama to win (M = 1.00, SD = 5.87) was not significantly different from those who wanted Romney to win (M = −0.70, SD = 5.79), t(57) = 0.97, p = .335, d = 0.29.

³ When analysed separately, both political knowledge questionnaires were negatively correlated with the accuracy of foresight judgments in the memory condition, McGlone et al.: r(69) = −.29, p = .015; Pew: r(69) = −.34, p = .004.

⁴ When analysed separately, neither political knowledge questionnaire was significantly correlated with hindsight bias in the hypothetical condition, McGlone et al.: r(82) = .09, p = .401; Pew: r(82) = .09, p = .426.

⁵ The hindsight index for each candidate did not significantly differ based on which candidate participants wanted to win the election in the hypothetical condition: the hindsight index for Obama among those who wanted Obama to win (n = 52; M = 0.48, SD = 4.56) was not significantly different from those who wanted Romney to win (n = 20; M = 1.15, SD = 4.33), t(70) = 0.58, p = .575, d = 0.15; the hindsight index for Romney among those who wanted Obama to win (M = 0.74, SD = 5.13) was not significantly different from those who wanted Romney to win (M = 1.80, SD = 4.70), t(70) = 0.42, p = .423, d = 0.22.