Replication Data for: Measuring Perceived Skin Color: Spillover Effects and Likert-type Scales

Discrimination based on skin color, referred to as colorism, has been documented as a con- siderable problem in social science research. Most of this research relies on Likert-type ratings of skin color. For example, the widely used “Massey Martin Scale” (MMS) requires coders to rate subjects on a scale from 1-10, based on the similarity between the subject’s skin tone and ten shades of skin color on a palette. Some scholars have raised questions about measurement error in Likert-type skin color scales. It’s been shown, for example, that Black and White coders apply the MMS differently. We hypothesize that the coding of a person’s skin color will vary depending on the race of persons previously coded. To test this hypothesis, we conducted an experiment using a convenience sample of Mturk workers as coders. We find that the MMS is vulnerable to spillover effects: a person’s skin is coded as “darker,” on average, if he is ob- served following a sequence of White persons than if he is observed following a sequence of Black persons. We also replicate previous work showing that Black and White coders use the scale differently. Finally, having coders cross-reference the palette at the time of coding, rather than recalling the palette from memory, fails to mitigate either race-of-coder or spillover effects. We provide suggestive evidence that use of a pairwise-comparisons approach may overcome some of the issues associated with Likert-type ratings of skin color.