Music preference and socioeconomic condition : evidence from QQ-music chart

Abstract

The success and popularity of a song are influenced by a multitude of factors. This article delves into the relationship between music preference and the socioeconomic condition in China. Advanced computer science tools enable extracting information from audio signals and textual content of a song. This study employs acoustic fingerprints obtained from Spotify's intelligence data platform and text features extracted using natural language processing (NLP) methods, including a knowledge-based representation method, a topic modeling method, and a pre-trained language model (BERT). The study uses the acoustic and lyrical content features of QQ-music's top 100 songs each week from September 2020 to June 2023 to predict the songs' position under different economic conditions in China. Each predictive model measures the relative importance of lyric features and acoustic features. The findings indicate that lyric features are more critical when the economic condition is robust, whereas acoustic features have more significance when the financial condition is weak. Furthermore, the cognitive resources of individuals are more focused on music during periods of relatively good economic conditions, leading to the attentive recognition of song lyrics. This is because speech recognition while listening to music is not essential and requires cognitive resources.