S.I.4. Robustness analyses measuring nomination success as position in top 30 percent
Below we change the operationalization of ballot position from the measure used in the main analyses (if a candidate received a position in the top 20 percent of his party’s ballot (coded “1”) or not (coded “0”)) such that candidates positioned in the top 30 percent of the ballot are now coded “1”, while candidates positioned lower than the top 30 percent are coded “0”. Table S.I.3 reports models corresponding to the models reported for the “top 20 percent variable” in the main text and Figure S.I.4.A and Figure S.I.4.B show predicted probabilities for receiving a top ballot position for candidates low (10th percentile) and high (90th percentile) in facial competence and facial dominance, respectively. Figure S.I.4.B further report separate results for liberal and conservative candidates. Importantly, all main conclusions replicate when using the top 30 percent measure as operationalization for nomination success.
Table S.I.4: Prediction of candidate ballot position in the top 30 percent from facial traits. Model 1 reports effects for facial competence, facial dominance and the interaction between facial dominance and candidate ideology. Model 2 further includes the interaction between facial competence and candidate ideology. Models report unstandardized logit regression coefficients with standard errors in parentheses.
Figure S.I.4.A: Predicted probabilities for position in top 30 percent of the ballot for candidates low (10th percentile) and high (90th percentile) in facial competence. Dashed lines are 95 percent confidence intervals.
Figure S.I.4.B: Predicted probabilities for position in top 30 percent of the ballot for candidates low (10th percentile) and high (90th percentile) in facial dominance from liberal (left-wing panel) and conservative parties (rightwing panel), respectively. Dashed lines are 95 percent confidence intervals.