Cultural Divergence in Popular Music: The Increasing Diversity of Music Consumption on Spotify Across Countries

The study examines whether digitized music consumption is leading to cultural convergence or divergence across countries. Digitization and streaming reduce marginal costs, expand catalogs, and lower consumer search and access costs, potentially increasing diversity. Conversely, easier access to international music could fuel convergence and winner-takes-all dynamics. Prior findings are mixed due to limited country samples and data granularity. Platform processes (e.g., recommendation algorithms, playlist curation) may confound observed preferences, and user demographics on platforms like Spotify are not fully representative. To address these concerns, the authors compare Spotify and iTunes charts and restrict analysis to 39 countries where Spotify is well established. They measure cross-national diversity over time to determine whether popular music charts are converging or diverging and use the Rao-Stirling diversity framework to capture variety, balance, and disparity of songs, artists, and labels, complemented by Zeta diversity to assess the prevalence of local versus global hits.

The research background contrasts two perspectives. Winner-takes-all theories highlight highly skewed success distributions in cultural markets, where scalability and path dependence can concentrate attention on a few global hits, increasing market concentration among hit songs, superstar artists, and major labels. In contrast, the Long Tail thesis argues that online retailing and distribution remove shelf-space constraints, increase catalog size, and enable niche consumption, boosting diversity and production via lower costs and new technologies. Recent extensions (Aguiar and Waldfogel) emphasize unpredictability in cultural markets, proposing a random long tail: digitization allows production of previously overlooked works, some of which achieve unexpected popularity; independent labels gain ground (e.g., US indie top-selling albums rose from 12% in 2000 to 35% in 2010). Together, these views frame a debate on whether digitization increases diversity or amplifies concentration. The authors position their study within this debate to empirically assess trends in cross-national diversity of popular music consumption.

The study analyzes national daily top charts from two platforms. Data: Spotify Top 200 daily charts across 39 countries from 2017-01-01 to 2020-06-20; iTunes Top 100 daily charts (19 countries; used for validation) from 2013-08-14 to 2020-07-16 (sourced via Kworb.com). For cross-platform comparison of variety, the authors also compute metrics on a common subset of 16 overlapping countries and align chart sizes at 100 positions, smoothing daily series over a 10-day window. To mitigate platform-specific confounds and user-base instability, they focus on countries where Spotify is strongly established (reducing from 59 to 39). Measures of diversity follow the Rao-Stirling framework (Stirling, 2007): - Variety: number of distinct units (songs, artists, or labels) divided by total chart positions. - Balance: 1 − Gini of the distribution of chart positions across units (higher indicates more even distribution). - Disparity: Euclidean distances based on acoustic features of songs (Spotify’s features: danceability, energy, key, loudness, mode, speechiness, acousticness, instrumentalness, liveness, valence, tempo, duration). Artist disparity is computed from the central tendency of the features of their charting songs. Rao-Stirling diversity combines variety, balance, and disparity via D = Σ_i Σ_j d_ij p_i p_j, with p_i, p_j as proportions and d_ij pairwise distances. To capture the prevalence of local versus global hits, the authors employ Zeta diversity (ζ) from ecology, computing the expected number (or percentage) of songs shared by groups of k countries for orders k=2 up to 20. Higher ζ-order values reflect more global hits; analysis tracks monthly evolution and the decay of ζ across orders. Additional analyses include: - Jaccard distances on annually constructed incidence matrices of songs across countries, visualized via multidimensional scaling to assess geographic/cultural clustering. - Two-mode country–song network projections with modularity to quantify clustering by region/culture over time. - Stratification of charts (Top 10, Top 50, Top 200) to assess whether diversity changes differ at the head versus the tail of charts. Statistical modeling: linear regression of log10(Z-diversity) on log10(Z-order), time (Month), and their interaction to assess changes in steepness over time (Table 1).

- Cross-national diversity in Spotify charts increased from 2017 to 2020 across songs, artists, and labels. Variety, balance, and disparity all rose, with a plateau in early 2020. Disparity exhibits a strong seasonal component around Christmas. - iTunes charts display a similar upward trend in variety over time, closely mirroring Spotify despite platform differences, supporting external validity. The increase on Spotify begins in 2017, plateaus by end of 2019; iTunes continues rising. - Countries cluster geographically/culturally in song-sharing space; modularity of the country–song network increases from 2017 to 2020, indicating consolidation of regional clusters that are more similar internally and more distinct from others. - Artist diversity increased across all components. Balance and variety increased more at the head of charts (Top 10) than at lower ranks, but artist disparity is lowest within the Top 10, suggesting stylistic similarity among peak-charting artists. - Label market structure became more even (higher balance) and more diverse (higher variety), with steeper increases at the head of charts. Labels had, on average, fewer songs and artists on charts in early 2020 than in early 2017: average songs per label decreased from 5.88 (2017 H1) to ~4.88 (2020 H1), and artists per label from 2.19 to 1.66; meanwhile, songs per artist increased slightly from 2.67 to 2.96 (H1 2017 vs H1 2020). - Zeta diversity analyses show declining mean overlap across countries over time at all orders and an increasingly steep decay with order, indicating that truly global hits became rarer relative to local/regional hits. Regression on log10(Z-diversity) yields significant negative effects of log10(Z-order) (−0.030, SE 0.001, p<0.01), Month (−0.009, SE 0.0004, p<0.01), and their interaction (−0.0001, SE 0.0001, p<0.05); R^2=0.684. - Overall, popular music charts have diverged across countries (Cultural Divergence), with the strongest increases in diversity occurring at chart peaks.

The findings address the core question by showing that digitized music consumption has not led to global convergence of popular music charts; instead, cross-country diversity increased from 2017 to 2020, consistent with Cultural Divergence. The rise in diversity is driven by segmentation: countries within culturally/geographically proximate clusters become more similar to each other while diverging from other clusters. Increased label diversity suggests that diversification is not merely major labels tailoring glocalized products, but involves broader changes in production/distribution. Zeta results indicate that global hits are increasingly rare, while local and regional successes dominate, contradicting pure winner-takes-all expectations under digitization. The authors argue that rapid, ubiquitous shifts are more plausibly explained by production-side dynamics (a random long tail enabled by digitization) than by large-scale, rapid changes in consumer preferences alone. While increased diversity may be positive for cultural distinctiveness, market segmentation could pose concerns if it extends from aesthetic to normative domains.

This study documents a pronounced rise in cross-country diversity of popular music consumption on Spotify (2017–2020), validated with iTunes, across songs, artists, and labels. Diversity increases were strongest at chart peaks, and countries clustered more distinctly by geography/culture over time. Applying Rao-Stirling and Zeta diversity, the authors show that global hits became rarer relative to local/regional successes, characterizing a process of Cultural Divergence rather than convergence or winner-takes-all globalization. Contributions include: (1) large-scale, multi-country, multi-platform evidence of increasing diversity; (2) extension of diversity analysis from songs to artists and labels; (3) methodological integration of ecology-inspired Zeta diversity to capture multi-country overlap dynamics. Open questions remain about mechanisms: whether changes are primarily driven by user preferences or by shifts in production and distribution (e.g., random long tail). Future work could disentangle these mechanisms and assess longer-term post-2020 trends as platforms and recommendation systems evolve.

- Platform confounding and non-independence: Spotify and iTunes may be influenced by similar industry-wide changes (e.g., recommendation systems, catalog expansions), and platforms may mutually influence each other. - User representativeness: Spotify users skew younger and more male than national populations; adoption timing correlates with listening habits (late adopters show stronger local preferences). - Country selection: Analysis restricted to 39 countries where Spotify is established to reduce instability; results may not generalize to newer markets (e.g., rapidly expanding markets like India). - Algorithmic and curation effects: Changes to recommendation systems (e.g., Spotify in 2019) and playlist practices (e.g., New Music Friday with national bias) could affect observed diversity. - Data sourcing: iTunes data obtained via a third-party aggregator (Kworb.com), not directly from an official API. - Time window: Spotify analyses cover 2017–mid-2020; trends plateau near 2020, and longer-term dynamics beyond this period are not assessed. - Charts-based view: Metrics reflect chart composition, which may be influenced by promotional dynamics and platform-specific factors beyond pure consumer preference.