Price Differentials and Skin Tone in Digital Art

Abstract

Non-fungible tokens (NFTs) are unique digital assets that verify ownership of data, enabling the creation of markets for digital artwork.  One of the pioneering NFT collections is CryptoPunks, which consists of 10,000 digital art images, most of which resemble human faces. Each 24 by 24 pixel image is a random collection of possible facial features and accessories, creating a lab-like environment to study how different attributes impact an artwork's auction value.  Using the entire collection transaction history of bids, asks, and sales that cover periods with median sales values between \$100 and \$400,000, we document a disturbing trend in how different features are rewarded in the market. As expected, scarcity is valued---except when it comes to skin tone, for which lighter skinned Punks sell for significantly more than darker skinned Punks despite being more common. This price gap, which emerged as early as 2018, has persisted through periods of immense growth in the market, leading to price differences of tens of thousands of dollars for otherwise similar images.  Our findings are robust across various empirical models and highlight the persistence of real-world biases in emerging digital markets. 

Keywords: Skin tone bias, digital economies, art valuation, non-fungible token