Intro: Print Music and Sleepify. 

For 400 years, music publishing was dominated by sheet music (printed notes and lyrics). The earliest known sheet music was a set of liturgical chants published in 1465 shortly after the Gutenberg Bible. Before the printing press, a composer's works had limited geographical reach, with music access limited to moneyed aristocrats. However, through mechanized music printing, an artist's work could be archived and distributed at relatively low cost to consumers worldwide. The printing press would be the first technology to disrupt the music industry.   

Flash forward to 2014. Spotify and iTunes exist, while sheet music is used exclusively by band geeks and symphony orchestras. The Michigan-based funk band Vulfpeck releases Sleepify, a ten-track album devoid of audible sound. Each track runs 30 seconds long and consists solely of silence. Sleepify is distributed via Spotify, where the band encourages fans to play the album on a loop while they sleep.

At the time of Sleepify's publication, Spotify paid a rate of $0.0030 to $0.0038 per stream lasting 30 seconds or more. In total, Vulfpeck's album received 5.5 million track listens for a tidy profit of $19,655.56. Somehow three hundred seconds of silence became a rousing commercial success — as far as streaming economics were concerned.      

Since the time of printed compositions, music publishing has seen myriad mediums come and go: live performance, radio, vinyl, cassette tapes, music videos, CDs, Napster, Limewire, iTunes, Pandora, Spotify, Youtube, Apple Music, and TikTok. With each change in distribution, artists adapt their acts to reflect shifting consumer preferences. So how has popular music changed throughout the years? And how has listening format impacted the content produced by musicians?

Methodology: How Has Popular Music Changed? 

Our goal is to track changes in song composition for popular music. We'll define popular songs as works listed on "The Billboard Hot 100" charts and utilize Spotify's database of song attributes to track changes in music design.

Dataset

Spotify parses audio feature information for every track on its platform. The company utilizes the resulting datasets for training recommendation algorithms and designing niche-specific playlists. 

Our analysis makes use of the following features from Spotify's dataset:  

• Song duration: track duration in minutes.

• Danceability: Describes how suitable a track is for dancing based on a combination of musical elements. A value of 0.0 is least danceable and 1.0 is most danceable.

• Instrumentalness: The closer the instrumentalness value is to 1.0, the greater likelihood the track has no vocal content.

• Speechiness: The more exclusively speech-like the recording (e.g. talk show, audiobook), the closer to 1.0 the attribute value.

• Valence: Measurement from 0.0 to 1.0 describing the musical positiveness conveyed by a track. Songs with high valence scores are more optimistic.

We will examine changes in the abovementioned song attributes for works listed on Billboard's Hot 100 Charts. 

The Billboard Hot 100 is the de facto record chart for the US music industry. Billboard's rankings derive from sales (physical and digital), radio plays, and online streaming traffic. Our dataset features all Billboard charts released between 1958 and late 2021.

Analysis Questions:

Music trends are fluid. A song's popularity is the product of ever-shifting cultural tastes, technological innovations, fleeting subcultures, and societal norms. We'll explore the music industry's frenetic evolution by examining the following questions: 

• How is popular music changing in length? 

• Is music popular evolving into a winner-take-all market?

• Is popular music becoming more or less dance-y?

• Is popular music changing in vocal and instrumental content?

• Is popular music growing more or less negative?    

Is Popular Music Changing in Length?

In the late 1950s, the 45-rpm (revolutions per minute) record format overtook the inelegant 78-rpm design, expanding average music capacity from 4–5 minutes per vinyl side to 9–12 minutes. This increase in revolution time would be the first of many innovations in data engineering. The second half of the twentieth century saw rapid advancement in form factor and storage capacity as album design progressed from vinyl to cloud storage. As time constraints disappeared, average track lengths increased, with popular song duration peaking in the mid-90s. 

Looking at artists with hit songs of above-average duration, it's clear the 80s and early 90s were a time of unconstrained creative freedom (for track duration, at least). 

Streaming was a distant threat, records and cassettes were the predominant format for distribution, and music videos encouraged in-depth storytelling alongside hit songs. Everybody wanted their MTV and didn't seem to care how long music lasted if played alongside pleasing visuals.  

And then, starting in the late 1990s, song length began to dip. There are numerous theories as to the shortening of popular music. The first (and laziest) theory is that consumer attention spans fell. That may be true, but shorter attention spans are usually an output rather than a causal factor (you can't just say people are stupid because they're stupid).

Likely the fall in song duration stems from two paradigmatic shifts in music economics. First, the introduction of iTunes and streaming led to a decoupling of songs and albums. Artists began focusing on the commercial maximization of individual tracks over cohesive album concepts. Second, streaming services typically pay per play, rewarding artists for the number of listens as opposed to total listening time. Streaming ten 31-second tracks pays more than a single stream of a 310-second song.   

Is Music A Winner-Take-All Market?             

As a marketplace, the music business connects listeners with free time to artists who wish to occupy that free time. Musicians fight for consumer mind share, much like any other media business. Until recently, record labels were an overwhelming determinant of artist reach, as a cartel of brands functioned as gatekeepers for physical music distribution (CDs and vinyls). And then Spotify came along. 

Spotify and musicians exist in an unhappy marriage marked by asymmetric power dynamics. Spotify emphasizes its role as a democratizing force within the music industry, providing consumers with ease and artists with decreased barriers to entry. I always assumed this last point was a means of obfuscating streaming's negative impact on artist compensation. But it appears the democratization argument holds some weight:

Since Billboard's inception, artists and labels have continuously maximized average chart tenure for hit songs. From the 1950s to the mid-2010s, a successful single remained in the public consciousness (and on the charts) for longer periods, while the number of artists listed on the charts remained mostly flat. In the middle of the 2010s, when streaming began its ascent, we see a trend break in chart longevity and quantity of artists listed. More artists make the charts for shorter periods. Maybe Spotify isn't all that bad (for musicians). 

Is Music More or Less Dance-y?

What makes a song danceable? People have cut loose for ages, regardless of Spotify's danciness ranking. It's not like people in the late 1950s abstained from dance because they were holding out for Outkast or Lil Baby. No, they listened to Bobby Darin and Elvis and thought, "wow, I love dancing to this sufficiently dance-able music."

And yet, according to Spotify, popular music has increased in danciness over time: 

So what is driving this purported uptick? A few theories:

1. Recency Bias in Spotify's Algorithm: Popular dance music typically exists within distinct cultural epochs (50s Doo-Wop, 60s Motown, 70s Disco, etc.). Would you expect the same playlist at a wedding and a nightclub? No. Nightclubs are for the young and promote dance by playing electronic music, hip hop, and rap, while weddings serve a broad age range, often catering to older demographics by playing from a list of well-worn classics (like "September" by Earth, Wind, and Fire). Spotify's algorithm may assess song attributes based on its (more youthful) base's dance preferences.   

2. The Importance of Live Music: The digitization of music distribution offers consumers abundant music content in return for a tidy and affordable subscription fee. In response, music acts shifted their profit centers from record sales to live performances. And what do people like to do at music shows? Dance (while wearing a flower crown)! Hip-hop, rap, and electronic music are cheap to produce and easy to promote, thus fetching increased attention from record labels and concert promoters.            

3. Industrialization of Music Production: In 1998, Cher's mega-hit single "Believe" made prominent (and ear-punishing) use of auto-tune. The computerized effect was a central feature of the song and served as a watershed moment for the mechanization of music production. Since that time, song creation has grown increasingly computerized, with major studios adapting digital workflows and programs like Logic and GarageBand democratizing prosumer production. Artists have slowly converged on a homogenetic sound for hip-hop, rap, and electronic music, with music technology capable of producing optimally dance-able beats.

4. The Rise of Hip-hop and Rap: Sorting through our artists repeatedly featured on the billboard charts, it appears Spotify's algorithm categorizes modern hip-hop and rap music as the most danceable (chart 1 below), while classifying traditional pop (from the 50s and 60s) and soft rock as least danceable (chart 2 below).   

Where Did the Instrumentals Go?

There is a time-honored tradition in rock DJ-ing known as the "toilet track," an established staple of songs long enough to allow the DJ a trip to the bathroom. Lynard Skynard's "Freebird," Queen's "Bohemian Rhapsody," and Led Zeppelin's "Stairway to Heaven" are prototypical examples of high-quality toilet tracks with extended guitar solos. But the toilet track may be a thing of the past, as Spotify's dataset observes a decrease in the prevalence of unaccompanied instrumentals, concurrent with an increase in lyric-driven music.

Zeroing in on artists with notable speech and instrumental content, we see the former category mainly associated with rap artists (chart 1) and the latter with guitar-forward rock and roll acts (chart 2). 

Perhaps it's a stretch, but the fall of instrumental music underscores rock's declining influence, and the rise of speechiness emphasizes rap's growing cultural dominance. Rock and roll is dead. Long live rock and roll.   

How Have Music Vibes Changed?

One temptation when reading the tea leaves of our Spotify data is to say that everything sucks. Music used to work, and now it's broken — what a bummer. The continuous decline in song positivity provides yet another tantalizing datapoint in support of music nihilism:

But I don't interpret this chart as evidence of music's darkening worldview. Instead, my hot take is that modern music better reflects our complicated everyday experiences. Consider our list of famous artists deemed highly positive by Spotify: 

We can split this roster of musicians into two camps, artists in 1950s America and 1970s dance jams. 

1. 1950s America was the age of Leave it to Beaver, I Love Lucy, the Jim Crow South, TV couples sleeping in separate beds, and Joseph McCarthy's red scare. Media was overly-positive, prohibited from addressing complex emotions or social phenomena, with explicit content facing a militant ban from the airwaves.   

2. 1970s dance jams are a relic from the days of disco. These songs induce euphoria, not profound introspection. I have never listened to Donna Summer or Kool & The Gang and thought, "this song is giving me feelz," or "I now possess a nuanced understanding of the human condition." No, I typically listen to these bands and think, "wow, this DJ is crushing this Bar Mitzvah."  

As means of comparison, consider Blink 182's "Adam Song," which represents a stark emotional departure from the abovementioned music. Released in 2000, "Adam's Song" takes the form of a fictional suicide note and was intended to inspire optimism for those struggling with depression. Many with mental health struggles have cited a profound connection with the song and its ultimate message of hope. I highly doubt "Adam's Song" would make it to the airwaves if released even a few years earlier.     

Final Thoughts: The Ever-Changing Medium and Ever-Responding Message.        

Canadian philosopher Marshall McLuhan is well-known for prophesizing a fragmented media landscape defined by ever-changing advances in distribution. His theories read as inevitable when evaluated against our current world of content abundance. McLuhan, however, was writing in the 1960s, when people bought Elvis records, watched four TV channels, and knew nothing of Elon Musk. Simpler times. 

In his book Understanding Media, McLuhan highlights the transformative power of technological progress (which he broadly terms "media" or "medium"):  

According to McLuhan, technological progress "shapes and controls" human interaction, not any one piece of content distributed via these new-fangled platforms. So, for example, it's the advent of social media that spurs change in human relations, not any one tweet that follows. 

Streaming offers a world of democratized music distribution and consumer abundance. Musicians can easily disseminate works in a market characterized by increased artistic control, and for $9.99 a month users get access to every song in human history. Today’s music industry, from songwriting to album promotion, is thus organized in service of a recurring $10 price commitment.    

And with that, let's return to Sleepify — sweet, wonderful, Sleepify. While overwhelmingly bare in content (or lack thereof), Sleepify’s success epitomizes (and mocks) music's maniacal quest to maximize time consumption. In this extreme case, a silent album generated nearly $20k in earnings but would have sold zero copies of sheet music in the late 1400s. Oh, how times have changed since Renaissance Italy.

Vulfpeck's muted masterpiece is uncomplicated yet far-reaching in its message, as the album required minimal production, simultaneously critiqued and exploited Spotify's business model, and made decent money in the process. It's also an example of artists acknowledging the regrettable business dynamics of streaming while simultaneously adjusting to the new norm. 

Streaming is here to stay — until the next disruption in music distribution — the artists that adapt their message will conquer the new medium. 

Want to chat about data and statistics? Have an interesting data project? Just want to say hi? Email [email protected]