Decision-makers frequently evaluate others under incomplete information. And a growing literature shows that evaluators rely on salient but non-diagnostic cues even in cases where “real” information is widely accessible, such as in the choice of CEO’s. We study the role of voice pitch—a seemingly economically irrelevant but perceptually salient trait—in performance prediction and hiring decisions.

Using a two-stage incentivized experiment, we first collect objective performance, and self-confidence information, as well as voice recordings from workers completing a real-effort task. In a second stage, independent evaluators (employers) predict performance and make hiring decisions based solely on workers’ voices. Voice pitch is manipulated within each speaker, allowing causal identification.

We find strong voice-pitch bias: lower-pitched voices are predicted to perform better and are more likely to be hired, despite no performance advantage. We additionally find gender-bias: men are predicted to perform better than women. Providing evaluators with feedback about realized performance substantially attenuates pitch bias in predictions but does not eliminate gender bias. Our findings demonstrate that seemingly economically irrelevant vocal cues distort evaluation and selection even in incentivized environments, leading to systematic misallocation.