The purpose of this study was to explore the effects of the numeracy and educational
levels on biases in five decision tasks: denominator neglect, risk-format effect in breast
cancer screening risk communication, misunderstanding of correlation, framing effect,
and conjunction fallacy. An internet survey was conducted with Japanese citizens in
metropolitan areas (n = 960) whose numeracy scores were previously measured by the
Japanese version of Lipkus et al.’s (2001) numeracy scale, aged from their 20s to their
60s with high and low educational levels. Data were analyzed based on 3 criteria of
numeracy levels: median split (10), split according to the previous study’s criterion
(9), and top-quartile (11) vs. bottom-quartile (7 and under). In the results of analysis
by median split, there were no significant differences except the denominator neglect.
There existed significant differences in educational levels across the three tasks. In the
results by the other criteria, there were significant differences in the tasks of conjunction
fallacy and framing effect, but those biases were rather stronger in the high-numeracy
group. The influence of the ceiling effect measuring numeracy in Japanese citizens
compared to the previous studies’ participants was discussed, as were the kinds of tasks
and the difference of response by numeracy.