Measuring National Happiness with Music

Emmanouil Benetos, Alessandro Ragano, Daniel Sgroi, Anthony TuckwellWarwick 2021

Abstract

We propose a new measure for national happiness based on the emotional content of a country’s most popular songs. Using machine learning to detect the valence of the UK’s chart-topping song of each year since the 1970s, we find that it reliably predicts the leading survey-based measure of life satisfaction. Moreover, we find that music valence is better able to predict life satisfaction than a recently-proposed measure of happiness based on the valence of words in books (Hills et al., 2019). Our results have implications for the role of music in society, and at the same time validate a new use of music as a measure of public sentiment.

Introduction

Quantifying the collective happiness of a nation is a fundamental challenge for economists and policymakers. Traditional metrics, such as the Eurobarometer survey, provide high-quality data on subjective wellbeing but are resource-intensive and often lack historical depth. Recent advancements in sentiment analysis have utilized large-scale text corpora, such as digitized books, to proxy national mood. However, literature often reflects long-term cultural shifts rather than immediate emotional responses.

This paper introduces the Music Valence Index (MVI), exploring whether the "language of emotions"—music—can serve as a more responsive and accurate mirror of public sentiment. By analyzing the "valence" (the intrinsic pleasantness or unpleasantness) of top-charting hits, we investigate the potential of popular music to act as a cultural thermostat, reflecting the collective emotional state of the population.

Methodology

The construction of the MVI involved a sophisticated machine learning pipeline designed to translate acoustic properties into emotional scores.

Data Selection and Feature Extraction

The study focused on the UK's #1 singles from 1973 to 2010. For each year, 45-second audio clips were processed using the Librosa and Essentia libraries. We extracted 191 distinct acoustic features, including Mel-Frequency Cepstrum Coefficients (MFCCs), Zero Crossing Rate, and Spectral Centroid, to capture the complex textures and moods of the music.

Model Training and Prediction

A Support Vector Regression (SVR) model was trained on the emoMusic dataset, which contains 744 songs annotated by human listeners for valence on a 9-point scale. This model achieved an $R^2$ of 0.33 on test data, demonstrating robust performance in automated emotion recognition. The trained model was then applied to the UK's annual top hits to generate a continuous index of musical valence over nearly four decades.

Results

Our analysis reveals a significant positive correlation ($r = 0.39, p = 0.02$) between the MVI and survey-based life satisfaction data from the Eurobarometer. This relationship holds true even after controlling for economic factors like GDP, inflation, and life expectancy.

Music vs. Literature

When compared against the Text Valence Index (TVI) derived from the Google Books corpus, the MVI proved to be a superior predictor. In multivariate "horse race" regressions, the MVI remained a statistically significant predictor of life satisfaction, while the TVI lost its predictive power. This suggests that music may be more uniquely attuned to the immediate emotional state of the populace than published literature, which can be a slower-moving medium.

The "Winner-Takes-All" Phenomenon

Interestingly, the predictive power of music is concentrated at the very top of the charts. The valence of the #1 song is a significantly better predictor of national happiness than the valence of the #2 or #3 songs, or even an average of the Top 10. This implies that the most popular song captures a unique "cultural resonance" that lower-charting tracks do not, acting as a singular reflection of the national zeitgeist.

Discussion and Conclusion

The findings suggest that popular music is not merely entertainment but a reflection of the collective psyche. The MVI offers a cost-effective, real-time proxy for national wellbeing that complements traditional survey methods.

While this study establishes a strong correlation, several questions remain regarding causality: does happy music improve the national mood, or does a happy populace gravitate toward more positive sounds? Regardless of the causal link, the MVI stands as a validated tool for measuring public sentiment through the lens of popular culture. Future research should expand this methodology to other geographic regions to test the universality of music as a predictor of happiness.

Summary updated

Abstract

We propose a new measure for national happiness based on the emotional content of a country’s most popular songs. Using machine learning to detect the valence of the UK’s chart-topping song of each year since the 1970s, we find that it reliably predicts the leading survey-based measure of life satisfaction. Moreover, we find that music valence is better able to predict life satisfaction than a recently-proposed measure of happiness based on the valence of words in books (Hills et al., 2019). Our results have implications for the role of music in society, and at the same time validate a new use of music as a measure of public sentiment.
BibTeX
@inproceedings{benetos2021measuring,
  author = {Emmanouil Benetos and Alessandro Ragano and Daniel Sgroi and Anthony Tuckwell},
  title = {{Measuring National Happiness with Music}},
  booktitle = {Warwick},
  year = {2021},
}
Figure