Sentiment analysis of popular-music references to automobiles, 1950s to 2010s

Chenyang Wu,John Polak,Elizabeth Bengel,Scott Le Vine,Jason Czerwinski

doi:10.1007/s11116-021-10189-1

Abstract

In recent years, there has been a scholarly debate regarding the decrease in automobile-related mobility indicators (car ownership, driving license holding, VMT, etc.). Broadly speaking, two theories have been put forward to explain this trend: (1) economic factors whose impacts are well-understood in principle, but whose occurrence among young adults as a demographic sub-group had been overlooked, and (2) less well-understood shifts in cultural mores, values and sentiment towards the automobile. This second theory is devilishly difficult to study, due primarily to limitations in standard data resources such as the National Household Travel Survey and international peer datasets. In this study we first compiled a database of lyrics to popular music songs from 1956 to 2015 (defined by inclusion in the annual “top 40”), and subsequently identified references to automobiles within this corpus. We then evaluated whether there is support for theory #2 above within popular music, by looking at changes from the 1950s to the 2010s. We demonstrate that the frequency of references to automobility tended for many years to increase over time, however there has more recently been a decline after the late 2000s (decade). In terms of the sentiment of popular music lyrics that reference automobiles, our results are mixed as to whether the references are becoming increasingly positive or negative (machine analysis suggests increasing negativity, while human analysis did not find a significant association), however a consistent observation is that sentiment of automobile references have over time become more positive relative to sentiment of song lyrics overall. We also show that sentiment towards automobile references differs systematically by genre, e.g. automobile references within ‘Rock’ lyrics are in general more negative than similar references to cars in other music genres). The data generated on this project have been archived and made available open access for use by future researchers; details are in the full paper.

Highlights

There are two sharply distinct theories to explain the reversal of the long-term growth trends
‘Peak Car’) in young people’s car-related mobility indicators observed in many high-income countries beginning in the 1990s/2000s (Blumenberg et al 2016; Ciari and Axhausen 2015; Kuhnimhof et al 2012)
The change of popular music and the reference to automobile is analyzed by descriptive analysis, whereas the sentiment towards cars is investigated by bivariate correlation and linear regression

Summary

Introduction

There are two sharply distinct theories to explain the reversal of the long-term growth trends One school of thought focuses on changing economic circumstances and external constraints on young people’s mobility This incorporates candidate explanators such as GDP per capita (Bastian et al 2016), declining workforce participation and income levels (Bayart et al 2020; Blumenberg et al 2016; Delbosc and Currie 2014a), increasing costs of owning and operating a car (Bastian et al 2016; Chatterjee et al 2018; Klein and Smart 2017), and the advent of mechanisms that have made acquiring a driving license more onerous, time-consuming, and/or expensive (Thigpen and Handy 2018). “Results” section presents the results, and “Conclusions” section summarizes and concludes the paper

Literature review

Key findings

Results

Conclusions

To identify each token