Consumer Preferences for Diversity : A Field Experiment in Product Design

This paper describes a randomized controlled field experiment designed to measure how individuals respond to racial and gender diversity in representations of certain archetypical occupations. We ask participants in two pools – a tech conferences and online – to evaluate their user experience with an image search engine tool and randomize them to see either diverse or non-diverse images. Subjects in the less diverse treatment view images that are predominantly white men for the high-status occupations (boss and professor) and predominantly women for the low-status occupations (nurse and clerk). In the more diverse treatment, subjects view image sets that contain a more equitable distribution of gender and race. We observe that diverse images result in significantly higher ratings across all participants and find no evidence of in-group bias in this context. However, women are disproportionately more dissatisfied with the lack of diversity in high-status occupations (boss and professor) than men are. For the low-status words (clerk and nurse), we find weaker treatment effects and no heterogeneity in the satisfaction ratings by gender or race. Free response qualitative data provide a potential explanation for the findings: while the extreme underrepresentation of women and minorities in the high-status, low-pay professions is salient, the equivalent underrepresentation of white men in the low-status, low-pay professions is less salient. Correcting the asymmetry in the way we promote diversity in high-status and low-status domains has important policy implications