Fig 1.
The total number of studies, categories, and signals included in our meta-analysis, along with our statistical estimation strategy.
The numbers indicate the total counts of studies, categories, and signals. For a detailed overview of signals by category for which raw data was obtained, refer to Table S1 in S3 File. Please note that the total number of studies is 21, as some are included in more than one category. Example signals are presented in the middle column (the resume). The data sources are shown on the right-hand side: hiring managers made callback decisions based on resumes (in blue). Separately, we collected warmth and competence ratings on prolific, where participants (in red) only saw the respective signal (indicated in yellow). Our estimation strategy is visualized in the grey box in the bottom right corner: we used the averages of warmth and competence ratings to predict the callback percentage.
Table 1.
Linear probability regressions of callback rates on principal components and social perception ratings.
Fig 2.
Warmth and competence ratings across names and their association with callback rates.
(A) Each scatterplot shows warmth and competence for each name in the sample one study (with the first author name at the top). The correlations between the two rating scales are strongly positive in all eight studies (Table S4 in S3 File). (B) Correlations between callback rates and PC1 and PC2 components associated with specific names (aggregating all studies). Data from different studies are identified by colors, with the legend shown in panel (C). The slope coefficients, shown in Table 1 are ,
. (C) Forest plot of confidence intervals for study-specific estimates of the correlation between callback rate and the first principal component PC1,
(callback, PC1). All correlations are positive. The pooled effect is
. (D) Scatter plots of name-specific warmth and competence ratings showing the structure of PC1 and PC2.
Fig 3.
Warmth and competence ratings across categories and their association with callback rates.
(A) Each scatterplot shows warmth and competence for each category-signal (with the category name at the top). The correlations between the two rating scales are strongly positive in all nine categories (Table S5 in S3 File). (B) Linear regression of PC1 on callback by category and study. Data from different studies are identified by colors, with the legend shown in the center of row three. (C) Scatter plots of category-specific warmth and competence ratings showing the structure of PC1 and PC2. (D) Meta-regression of PC1 and PC2 on callback, where each circle identifies a signal by study; the circle size indicates the assigned weight in the meta-regression. Lines indicate fitted intercepts and .