Emergence of Long-Term Memory in Popularity

Hyungjoon Soh, Joo Hyung Hong, Jaeseung Jeong, Hawoong JeongarXiv 2017 · arXiv

Introduction

Why do some songs vanish in a week while others linger on the charts for months? Traditional models of population dynamics, such as the logistic equation, often fail to predict long-term popularity accurately because they assume a Markovian process where the future depends only on the present state. This research analyzes nearly 60 years of Billboard Hot 100 data to identify a fundamental shift in music popularity: the emergence of "long-term memory."

Modeling Popularity with Memory

The researchers proposed a model that integrates two distinct processes to describe a song's life cycle:

  1. Spreading Process ($S(t)$): A standard logistic growth function representing the initial "infection" or discovery rate of a song.
  2. Memory Kernel ($\Theta(s)$): A power-law decay function that weights past activity, allowing previous popularity to influence current success.

To map real-world data to the model, the authors used Zipf’s Law ($P(r) \sim r^{-1}$), assuming a song's popularity is inversely proportional to its Billboard rank. By fitting this model to 28,312 songs, they estimated parameters such as total potential popularity ($A$), growth time scale ($t_0$), initial fraction ($IF$), and memory length ($\tau_c$).

The Shift in Chart Dynamics: From News to History

A key insight from the study is the "death of the slow climb." In the early decades of the Billboard Hot 100 (1950s–1970s), songs typically followed a symmetric rise-and-fall pattern. Modern hits, however, frequently exhibit "Hot-Shot Debuts," entering the charts near the top and decaying very slowly.

Figure 1: Representative rank trajectories of modern hits, illustrating the "Hot-Shot Debut" and the persistent "long tail" of popularity.

The authors identify a "phase transition" in the early 1990s. Before this period, songs behaved like "news"—consumed and quickly forgotten. Afterward, they behaved like "history," where past popularity became a strong predictor of future rank. This shift is attributed to the transition from sales-based metrics (one-time purchases) to activity-based metrics (radio play, streaming, and views), which naturally accumulate and reflect the population's collective memory.

The model significantly outperformed standard null models, proving that a memory kernel is essential for describing modern popularity. The analysis revealed several long-term trends:

  • Increased Lifetimes: The average time a song spends on the chart has more than doubled since the 1960s.
  • Higher Initial Ranks: Songs now debut at significantly higher ranks (lower numerical values) than in previous eras.
  • Polarization: The "memory" effect creates a winner-take-all dynamic, where a small number of "long-memory" juggernauts dominate the charts for extended periods, increasing inequality in the industry.
Figure 2: Historical trends showing the increase in mean weeks on chart (left) and the shift toward higher initial ranks (right) over six decades.

Limitations and Future Work

While the model provides a robust framework for understanding popularity, it treats songs as independent events, ignoring the influence of an artist's previous success or "brand loyalty." Additionally, the model can be disrupted by exogenous shocks, such as an artist's death or a viral meme, which cause atypical rank paths. Future research could apply this "emergence of memory" framework to other ranking systems, such as book sales or social media trends, to determine if this is a universal feature of the digital attention economy.

Summary updated

Abstract

Popularity describes the dynamics of mass attention, and is a part of a broader class of population dynamics in ecology and social science literature. Studying accurate model of popularity is important for quantifying spreading of novelty, memes, and influences in human society. Although logistic equation and similar class of nonlinear differential equation formulates traditional population dynamics well, part of the deviation in long-term prediction is stated, yet fully understood. Recently, several studies hinted a long-term memory effect on popularity whose response function follows a power-law, especially that appears on online mass media such as YouTube, Twitter, or Amazon book sales. Here, we investigate the ranking of most popular music, \textit{the Billboard Hot 100 chart}, which is one of the largest popularity dataset spanning several decades. Using a popularity model that comprises logistic growth and a power-law decaying long-term memory, we showed that rank history is mainly characterized by initial popularity and memory strength. With this framework, we investigated temporal development of long-term memory on the whole popularity dynamics. As a result, abrupt emergence of long-term memory and broad initial popularity is illustrated, which was not clearly detected by time-independent measures. We emphasize not only development of the mass media, but also the difference of spreading and accumulated popularity affect dynamics significantly when the popularity has long-term memory.
BibTeX
@misc{soh2017emergence,
  author = {Hyungjoon Soh and Joo Hyung Hong and Jaeseung Jeong and Hawoong Jeong},
  title = {{Emergence of Long-Term Memory in Popularity}},
  year = {2017},
  eprint = {1712.02234},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1712.02234},
}
Figure