Incarceration Beliefs
indicating a significant difference from zero for each of the raw discrepancy scores. Based on the
standards described earlier, the average of the participants’ estimates of the proportion of the prison
population that were high school graduates and mentally ill were inaccurate; the other four estimates
were near misses.
Assessing consensual-level stereotype accuracy: Correlations. Next, we assessed how well
the consensual stereotypes corresponded with the actual demographics of the prison population. We
did this by correlating the estimated percentage (Column 3 from Table 2) with the actual percentages
of the population comprised by each group (Column 2 from Table 2). Because we are correlating
means (not individuals) with population percentages, and because there are only six such means, we
do not report a test of significance. Nonetheless, the correlation coefficient still provides a good
assessment of the extent to which the consensual stereotypes corresponded with reality. The
correlation of the consensual stereotype with the population percentages, r= .55, is accurate as per the
standards described in the introduction.
Assessing individual-level accuracy: Discrepancies. Results so far focused on consensual
stereotypes. The present section focuses on individual stereotypes. How many people’s judgments
were accurate, near misses, or inaccurate? The last three columns of Table 2 indicate the number of
participants whose estimates were within 10 percentage points (accurate), within 20 points (near
miss), and over 20 points (inaccurate). Participants appear to be the most accurate in their estimates
of Latinos in prison, and quite inaccurate in their estimates of high school graduates in prison. There
was a fairly even distribution of accurate estimates, near misses, and inaccurate estimates for the
other four social groups (females, the mentally ill, Blacks, and Whites).
Assessing individual-level stereotype accuracy: Correlations. At the individual level of
analysis, each participant’s estimated proportion was correlated with the actual proportion in each
demographic group (Column 2 from Table 2). In this manner, we computed an individual-level
16