Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Impact of Social Media on Music Demand: Evidence from Quasi-Natural Experiments

Daniel Winkler, Christian Hotz-Behofsits, Nils Wlömert, Dominik Papies, Jūra Liaukonytė

TL;DR

This paper analyzes whether TikTok complements or cannibalizes paid music streaming by exploiting two quasi-natural experiments: UMG’s 2024 withdrawal and the 2025 U.S. outage. Using a matched-pairs DiD framework and proportional-mean models (PPML and weighted log-OLS), it uncoveres strong heterogeneity: the viral head bears most of the off-platform demand losses when TikTok exposure falls, while long-tail songs show little response. The findings support a discovery-upstream and playlist-downstream amplification mechanism, with downstream consequences including Spotify playlist follower declines and reduced Top-200 chart presence for UMG. Methodologically, it shows that in power-law-like settings, different DiD specifications target different estimands, so explicit attention to estimand-Estimator alignment is crucial for interpreting aggregate effects. The results imply that TikTok enhances downstream streaming and rights holders’ visibility, affecting licensing negotiations and strategic decisions for multi-platform promotion.

Abstract

We study how TikTok affects demand for music on paid streaming platforms. We use two quasi-exogenous disruptions to TikTok access that generate independent variation in exposure: Universal Music Group's (UMG) global withdrawal of its catalog from TikTok in February 2024, and the 2025 U.S. TikTok outage and subsequent install restriction. Across both settings, we reach the same bottom-line conclusion: reductions in TikTok access lower off-platform streaming demand. Recent work using the UMG setting reaches mixed conclusions about whether TikTok primarily promotes or cannibalizes streaming demand. We show that these findings can be reconciled once we account for power-law-like concentration in exposure (10% of songs account for 95% of TikTok creations) and that, in such environments, common DiD implementations (levels, log, and Poisson) target different estimands and impose different counterfactual trend restrictions. We document sharp heterogeneity: the typical long-tail song exhibits little change when TikTok access is removed, while highly viral titles experience meaningful declines in Spotify streams. Because the viral head accounts for a disproportionate share of total listening and revenue, these losses drive the market-level implications. The evidence is consistent with a mechanism in which TikTok operates as an upstream source of discovery and Spotify playlists act as a downstream amplification layer: declines increase with TikTok virality, UMG songs with no prior TikTok exposure show null effects, and UMG songs accumulate fewer followers on Spotify curated playlists. We also document downstream consequences for rights holders, with UMG songs becoming less likely to reach top charts.

The Impact of Social Media on Music Demand: Evidence from Quasi-Natural Experiments

TL;DR

This paper analyzes whether TikTok complements or cannibalizes paid music streaming by exploiting two quasi-natural experiments: UMG’s 2024 withdrawal and the 2025 U.S. outage. Using a matched-pairs DiD framework and proportional-mean models (PPML and weighted log-OLS), it uncoveres strong heterogeneity: the viral head bears most of the off-platform demand losses when TikTok exposure falls, while long-tail songs show little response. The findings support a discovery-upstream and playlist-downstream amplification mechanism, with downstream consequences including Spotify playlist follower declines and reduced Top-200 chart presence for UMG. Methodologically, it shows that in power-law-like settings, different DiD specifications target different estimands, so explicit attention to estimand-Estimator alignment is crucial for interpreting aggregate effects. The results imply that TikTok enhances downstream streaming and rights holders’ visibility, affecting licensing negotiations and strategic decisions for multi-platform promotion.

Abstract

We study how TikTok affects demand for music on paid streaming platforms. We use two quasi-exogenous disruptions to TikTok access that generate independent variation in exposure: Universal Music Group's (UMG) global withdrawal of its catalog from TikTok in February 2024, and the 2025 U.S. TikTok outage and subsequent install restriction. Across both settings, we reach the same bottom-line conclusion: reductions in TikTok access lower off-platform streaming demand. Recent work using the UMG setting reaches mixed conclusions about whether TikTok primarily promotes or cannibalizes streaming demand. We show that these findings can be reconciled once we account for power-law-like concentration in exposure (10% of songs account for 95% of TikTok creations) and that, in such environments, common DiD implementations (levels, log, and Poisson) target different estimands and impose different counterfactual trend restrictions. We document sharp heterogeneity: the typical long-tail song exhibits little change when TikTok access is removed, while highly viral titles experience meaningful declines in Spotify streams. Because the viral head accounts for a disproportionate share of total listening and revenue, these losses drive the market-level implications. The evidence is consistent with a mechanism in which TikTok operates as an upstream source of discovery and Spotify playlists act as a downstream amplification layer: declines increase with TikTok virality, UMG songs with no prior TikTok exposure show null effects, and UMG songs accumulate fewer followers on Spotify curated playlists. We also document downstream consequences for rights holders, with UMG songs becoming less likely to reach top charts.
Paper Structure (53 sections, 25 equations, 17 figures, 16 tables)

This paper contains 53 sections, 25 equations, 17 figures, 16 tables.

Table of Contents

  1. Introduction
  2. Institutional Background and Data
  3. UMG’s TikTok Withdrawal and Re-entry
  4. Data
  5. Treated songs.
  6. Control songs.
  7. Power-Law Concentration in Spotify Streams and TikTok Creations
  8. Defining the viral head and the long tail.
  9. Model-free evidence by TikTok-virality decile
  10. Difference-in-Differences Diagnostics: Estimands, Estimators, and Interpretation
  11. Motivation and Terminology
  12. OLS in Levels.
  13. Log-OLS.
  14. Weighted log-OLS.
  15. PPML.
  16. ...and 38 more sections

Figures (17)

  • Figure 1: Weekly user-generated creations for UMG songs in the TikTok Top 100.
  • Figure 2: Concentration and Correlation Between TikTok Creations and Streams.
  • Figure 3: Joint Distribution of Spotify Streams and TikTok Creations Across Deciles
  • Figure 4: Weekly Mean Spotify Streams by Pre-Treatment TikTok-Virality Decile.
  • Figure 5: Treatment Effects by Pre-Treatment TikTok Virality.
  • ...and 12 more figures