Jose Oros

Playlist-Induced Spillovers: Examining On- and Off-Platform Outcomes for Emerging Artists on Spotify

with Jacqueline Doremus, Kyle Kretschman, Rahul Telang and Michael D. Smith

Job Market Paper

Digitization and streaming platforms-based activities can create opportunities for artists at all stages of their careers. This paper studies the extended multiplier effects of editorial playlisting on both on-platform and off-platform outcomes for emerging, independent artists. Utilizing a unique dataset from Spotify’s Playlist Pitching tool, a feature that allows artists to pitch their tracks to music editors for playlist consideration, we employ a matched difference-in-differences methodology to compare outcomes between artists who were featured on playlists and those who were not. Our results show that emerging independent artists featured on Spotify’s editorial playlists not only experience a significant increase in stream counts but also have a greater likelihood of going on tour. This research emphasizes the role of digital platforms as potent discovery mechanisms that enable fan engagement, in turn, driving off-platform career trajectories including live event opportunities.

Generative AI and Musical Creativity

with Richard Randall, Rahul Telang

Working Paper

Generative AI is swiftly transforming creative industries, particularly music, raising questions about its impact on artistic practices and artist roles. While tools like Suno and Udio, powered by large language models, can generate full songs from prompts, their impact on human musical creativity and quality remains unclear. This project aims to analyze the influence of generative AI on music creation, focusing on how AI assistance affects the creative process and artistic outcomes. We will conduct a randomized control trial with participants tasked with composing a 15-second melody, dividing them into two groups: one using AI tools for ideation and a control group without AI assistance. We will assess how both artists and audiences perceive AI-assisted compositions compared to human-generated ones. Additionally, the melodies will be analyzed across key dimensions, including creativity, musicality, and likeness. The project seeks to evaluate the implications of generative AI in artistic workflows and identify potential risks and opportunities. The findings will provide valuable insights for artists, platforms and policymakers.

Predicting Artist Success using Social Media Data

with George Chen, Michael D. Smith

Heinz College First Paper

Music labels have historically relied heavily on "gut feelings" to make decisions about which artists to sign. This is largely driven by the industry wide "Nobody-Knows-Anything'' mantra attributed to Hollywood producer William Goldman, describing the apparent impossibility of defining what content will be successful due to the high variance in results. In this research, we study how audience data extracted from social media can play a central role in informing decision makers and help mitigate the risk and uncertainty inherent to the artist scouting process. We employ machine learning and statistical methods to predict two different measures of success: whether the artist signs a deal to produce an album with a major record label and whether a new artist's work will appear in a Billboard chart. We train and test our models using data from 188,734 videos on YouTube, representing 6,259 artists that released music in the US in the period of 2017-2019.

Our classification results show that audience metrics from YouTube can be used to predict these two major milestones in an artist's career with some level of success. This finding contributes to the understanding of industry managers looking to increase the use of data in their business decisions, and the broader academic literature analyzing how social media data can be used to predict practical business outcomes.

Page updated

Google Sites

Report abuse