Unemployment Rate Predicts Anger in Popular Music Lyrics: EvidenceFrom Top 10 Songs in the United States and Germany From 1980 to 2017

Popular music has been shown to reflect cultural characteristics and psychological change in a society. However, little is known about how popular songs are related to the socioeconomic conditions. In this research, we analyzed the annual top 10 songs from United States and Germany between 1980 and 2017, and found that the unemployment rate predicted the amount of anger but not anxiety or sadness in lyrics in both countries. Our research contributes to the literature on popular media culture by revealing that top song lyrics may reflect public sentiment toward the socioeconomic environment. It highlights the possibility of using top song lyrics as an alternative measure of public sentiments.

Public Policy Relevance Statement
The unemployment rate in the United States and Germany predicted the amount of anger in the lyrics of each country’s annual top 10 songs. The findings suggest that popular music can provide a glimpse of public sentiment in response to the socioeconomic environment.

Keywords: song, music, lyrics, emotion, unemployment

 

Music is an important part of everyday life (Rentfrow, 2012; Rentfrow & Gosling, 2003). Popular songs in a country indicate what the majority likes and have been shown to reflect societallevel psychological and cultural changes in terms of individualism and self-promotion (DeWall, Pond, Campbell, & Twenge, 2011; McAuslan & Waung, 2018). Given that people like to listen to music that matches their mood and concerns (Chamorro-Premuzic & Furnham, 2007; Larson, 1995; Qiu, Chen, Ramsay, & Lu, 2019; Schwartz & Fouts, 2003), popular songs are likely to reflect the overall mood of a country. Because the socioeconomic environment can significantly influence people’s affect and well-being (see Dolan, Peasgood, & White, 2008 for a review), it is possible that sentiments in popular songs can reflect the collective mood in a country in response to the socioeconomic environment. Therefore, in this study, we focused on unemployment—one of the most severe problems affecting socioeconomic environments—and aimed to examine how sentiments (i.e., anger, anxiety, and sadness) in top song lyrics are related to the unemployment rate. To the best of our knowledge, this is the first study that proposes top song lyrics as a measure of societal-level emotional reactions to socioeconomic conditions. It provides new insights into the popular media culture and suggests that popular music may be utilized as a proxy of public sentiment toward the socioeconomic environment.

Music Choices Reflect Individual Concerns and Cultural Characteristics

Although songs in noninstrumental genres contain both melodies and lyrics, lyrics have been found to have unique effects beyond those exerted by melodies (Ali & Peynirciog˘lu, 2006; Anderson, Carnagey, & Eubanks, 2003). For example, Ali and Peynirciog˘lu (2006) measured participants’ affective reactions to melodies with or without lyrics and found that emotionally congruent lyrics reduced the positive emotion conveyed by happy and calm music but increased the negative emotion conveyed by sad and angry music. Anderson et al. (2003) asked participants to listen to tense music with either violent or nonviolent lyrics and found that the content of lyrics, rather than the tense rhythm or distorted sound, resulted in aggressive thoughts and hostile feelings. Furthermore, studies in psychophysics and neuroscience have shown that melodic and lyrical information are processed independently when people listen to music (Besson, Faita, Peretz, Bonnel, & Requin, 1998; Bonnel, Faita, Peretz, & Besson, 2001).

Previous studies have suggested that individuals are often drawn to lyrics that match their mood and concerns. For example, adolescents with few friends prefer songs with themes of loneliness or independence (Burke & Grinder, 1966), whereas those who enjoy heavy music (e.g., hard rock) tend to face issues such as low self-esteem, poor social connectedness, and an unstable sense of identity (Schwartz & Fouts, 2003). Moreover, European Americans with a college degree favor rock songs, whereas those without a college degree prefer country songs (Snibbe & Markus, 2005). These distinct preferences may be due to the specific types of values portrayed in different genres of music. Values commonly depicted in rock songs such as individual uniqueness and environmental control are likely to be endorsed by individuals with high socioeconomic status, whereas values frequently depicted in country music such as personal integrity and self-control are likely advocated by individuals with low socioeconomic status. Qiu et al. (2019) showed that individuals high in neuroticism like songs with few positive emotion words, which matches their low desire for positive emotion.

Popular songs in a country reflect what the majority likes and have been found to reveal societal-level values and norms (DeWall et al., 2011; McAuslan & Waung, 2018). Researchers have considered popular songs, together with books and TV advertisements, as cultural products that are developed by people to reflect and reinforce psychological processes in a particular sociocultural environment (Lamoreaux & Morling, 2012; Morling & Lamoreaux, 2008). For instance, cultural products from Western cultures have been found to be more individualistic and less collectivistic than their counterparts from Eastern cultures (Morling & Lamoreaux, 2008). In fact, a number of studies have shown that crosscultural differences in popular song lyrics reflect established cultural characteristics in a society. To illustrate, Rothbaum and Tsang (1998) demonstrated that popular love songs in China portrayed love as embedded within a larger context of a relationship as compared with American love songs. In addition, lyrics of popular songs in China contained more collectivist themes (e.g., expression of positive reciprocity toward one’s parents) than those in the United States (Rothbaum & Xu, 1995). Studies have also shown that longitudinal changes in popular song lyrics parallel cultural changes. Analyzing the top 10 songs ranked in the United States between 1980 and 2007, DeWall et al. (2011) found an increased usage of first-person singular pronouns (e.g., I, me) and a decreased usage of social process words (e.g., mate, talk) over the years. McAuslan and Waung (2018) extended these findings by showing that the top 100 songs in 2010 had significantly more self-promotion lyrics than those in 1990 and 2000. These two studies suggest that lyrics of top songs can reflect the rise of self-focus and individualism in the United States.

Top Song Lyrics and Socioeconomic Environments

Pettijohn and colleagues have examined trends in top songs during times of socioeconomic difficulty (Eastman & Pettijohn, 2019; Pettijohn, Eastman, & Richard, 2012; Pettijohn & Sacco, 2009a, 2009b). They used an aggregated index combining unemployment rate, disposable personal income, consumer price index, death rate, birthrate, marriage rate, divorce rate, suicide rate, and homicide rate to indicate the socioeconomic condition in the United States. Pettijohn and Sacco (2009a) analyzed Billboard No.1 songs from 1955 to 2003 and showed that longer sentences and words related to future and social processes appeared more frequently in lyrics during difficult socioeconomic times. Pettijohn and Sacco (2009b) further asked raters to listen to these songs and found that songs during threatening socioeconomic conditions were rated as longer in duration, more meaningful, more comforting, and more romantic. Pettijohn et al. (2012) examined the melodic attributes of Billboard No. 1 songs from 1955 to 2008 and showed that songs during social and economic bad times had less beats per minute and less common key signatures, suggesting that they were more reflective and serious. Finally, Eastman and Pettijohn (2019) analyzed Billboard rhythm and blues/hip-hop songs of the year between 1946 and 2010 and found that songs are more likely to be longer, slower, and about relationships with others rather than leisure or fun when socioeconomical conditions deteriorate.

While the aforementioned studies have shown important evidence of how popular music changes according to the socioeconomic environment, the scope of their data was limited as only the top one song for each year was analyzed. In addition, it is unclear as to which socioeconomic condition was actually related to the change in lyrics because several distinct socioeconomic indicators were combined. The unemployment rate has been shown to affect societal perception of the state of the socioeconomic environment (Bianchi, 2016; Hill, Rodeheffer, Griskevicius, Durante, & White, 2012), and predicts self-reported unhappiness above and beyond personal unemployment as well as inflation rate (Di Tella, MacCulloch, & Oswald, 2001; Wolfers, 2003). Moreover, it has a robust and strong negative relationship with subjective well-being across different time periods and populations (Dolan et al., 2008). Therefore, in this study, we focused on unemployment rate and disentangled its relationship with lyrics by controlling for other possible socioeconomic indicators.

Present Study

This study aimed to examine how sentiments in top songs coincide with changes in national unemployment rate. In particular, we focused on three common negative emotions (i.e., anxiety, sadness, and anger) expressed in lyrics. Previous research has shown that anxiety, sadness, and anger are related to different decision-making processes (Raghunathan & Pham, 1999), social information judgment (Bodenhausen, Sheppard, & Kramer, 1994), and emotion-regulation strategies (Blanchard-Fields & Coats, 2008). Although we expected that the negative sentiments in top songs were related to a country’s unemployment rate, we did not have specific hypotheses regarding how the three negative emotions would differ in their relationships to unemployment rate due to the lack of theoretical basis and empirical findings.

We used a widely used text analysis software program, the Linguistic Inquiry and Word Count (LIWC; Tausczik & Pennebaker, 2010), to compute the word frequencies in three predefined negative emotion categories: anger (e.g., hate, kill, annoyed), anxiety (e.g., worried, fearful, nervous), and sadness (e.g., crying, grief, sad). We considered top songs as a reflection of what the population likes to consume, regardless of where or when the song was produced. Therefore, we predicted that the relationship between sentiments in top songs and unemployment rate would hold in countries that are open to embracing foreign songs. Hence, we wanted to choose a country whose chart is dominated largely by locally produced songs (Sample 1) and a country whose chart has a significant number of foreign songs (Sample 2). Our selection was made by first considering countries whose official language has a corresponding LIWC dictionary. LIWC currently has dictionaries in the following languages: English, Spanish, French, Russian, Italian, Dutch, German, Brazilian Portuguese, Chinese, and Norwegian. After which, we identified countries that have an official top 10 music chart and common socioeconomic variables for over 30 years. Consequently, the United States became Sample 1 and Germany became Sample 2. Lyrics from the top 10 songs in the United States and Germany were gathered and analyzed using LIWC. We collected the national unemployment rate alongside other socioeconomic indicators (including gross domestic product [GDP] per capita, inflation rate, housing price, population density). Multiple regression models were conducted to examine the relationships between unemployment rate and three LIWC emotion categories (i.e., anger, sadness, and anxiety).

Method

The top 10 most popular songs between 1980 and 2017 in the United States and Germany were gathered, respectively.1 Subsequently, song lyrics were downloaded from http://www.metrolyrics .com/ or https://genius.com/. For the U.S. sample (see the online supplemental materials), a total of 379 lyrics (out of 380 songs) were analyzed because one instrumental song has no lyrics. Apart from two multilanguage songs (i.e., English and Spanish), all remaining songs are in English. For the German sample (see the online supplemental materials), a total of 366 lyrics (out of 380 songs) were analyzed, after removing 14 songs that are instrumental(n=5) or in languages (i.e., Latin, Hebrew, Duala, Hausa, Romanian, and Korean) that do not have a corresponding LIWC dictionary(n=9). The German sample contains lyrics in German(n=63), English(n=280), Spanish(n=5), Italian(n=5), French(n=4), and Portuguese(n=3). There are six multilanguage songs, including German and English(n=2), German and French(n=1), English and French(n=1), English and Italian(n=1), and English and Spanish(n=1). Their lyrics were first separated by language (e.g., English words and Grman words), and then each part was analyzed using its corresponding LIWC dictionary.

LIWC was used to compute word frequencies in three negative emotion categories (i.e., anger, sadness, and anxiety) in the lyrics. To illustrate, Kendrick Lamar’s hit song in 2017 hit song Humble contains a line “Piss out your per diem, you just gotta hate ’em, funk.” Because both piss and hate are in the anger category, the anger category will have a score of 2/11=.18. Previous studies have shown that LIWC can reliably measure emotional processes expressed in language samples (Pennebaker, Mehl, & Niederhoffer, 2003; Tausczik & Pennebaker, 2010; Tov, Ng, Lin, & Qiu, 2013) and have used LIWC to analyze lyrics. For example, Qiu et al. (2019) used LIWC to analyze the lyrics of participants’ favorite songs and revealed meaningful associations between personality traits and word use in these songs. DeWall et al. (2011) utilized LIWC to count the frequency of first-person singular pronouns (e.g., I, me), positive emotion words, and social process words in top song lyrics and suggested that frequency changes in these categories reflect changes in narcissism and individualism in the United States.

The U.S. employment rate was taken from the World Bank (2019c), and the German unemployment rate was taken from the Organisation for Economic Cooperation and Development (2019b). Previous work suggested a few other socioeconomic indicators that could influence emotional processes. For example, life satisfaction is related to the experience of positive emotions (Kuppens, Realo, & Diener, 2008), and GDP per capita has been found to be positively related to life satisfaction (Stevenson & Wolfers, 2008). Housing price has also been shown to be positively associated with life satisfaction because it indicates confidence in the economy (Ratcliffe, 2010). Both inflation (Di Tella et al., 2001) and population density (Li & Kanazawa, 2016) were found to be negatively related to life satisfaction. Therefore, these variables were included as control variables in the analyses. For both the United States and Germany, GDP per capita and population density were taken from the World Bank (2019a, 2009b), inflation was taken from Inflation.eu (2019), and housing price was taken from the Organisation for Economic Cooperation and Development (2019a).